chore: retire unmaintained workspace runtimes

README.md

@@ -53,7 +53,7 @@ Molecule AI is the most powerful way to govern an AI agent organization in produ
 It combines the parts that are usually scattered across demos, internal glue code, and framework-specific tooling into one product:
 
 - one org-native control plane for teams, roles, hierarchy, and lifecycle
-- one runtime layer that lets **eight** agent runtimes — LangGraph, DeepAgents, Claude Code, CrewAI, AutoGen, **Hermes**, **Gemini CLI**, and OpenClaw — run side by side behind one workspace contract
+- one runtime layer that lets **four** maintained agent runtimes — Claude Code, Codex, **Hermes**, and OpenClaw — run side by side behind one workspace contract
 - one memory model that keeps recall, sharing, and skill evolution aligned with organizational boundaries (Memory v2 backed by pgvector for semantic recall)
 - one operational surface for observing, pausing, restarting, inspecting, and improving live workspaces
 
@@ -75,7 +75,7 @@ You do not wire collaboration paths by hand. Hierarchy defines the default commu
 
 ### 3. Runtime choice stops being a dead-end decision
 
-LangGraph, DeepAgents, Claude Code, CrewAI, AutoGen, Hermes, Gemini CLI, and OpenClaw can all plug into the same workspace abstraction. Teams can standardize governance without forcing every group onto one runtime.
+Claude Code, Codex, Hermes, and OpenClaw can all plug into the same workspace abstraction. Teams can standardize governance without forcing every group onto one runtime.
 
 ### 4. Memory is treated like infrastructure
 
@@ -112,13 +112,9 @@ Molecule AI is not trying to replace the frameworks below. It is the system that
 
 | Runtime / architecture | Status in current repo | Native strength | What Molecule AI adds |
 |---|---|---|---|
-| **LangGraph** | Shipping on `main` | Graph control, tool use, Python extensibility | Canvas orchestration, hierarchy routing, A2A, memory scopes, operational lifecycle |
-| **DeepAgents** | Shipping on `main` | Deeper planning and decomposition | Same workspace contract, team topology, activity stream, restart behavior |
 | **Claude Code** | Shipping on `main` | Real coding workflows, CLI-native continuity | Secure workspace abstraction, A2A delegation, org boundaries, shared control plane |
-| **CrewAI** | Shipping on `main` | Role-based crews | Persistent workspace identity, policy consistency, shared canvas and registry |
-| **AutoGen** | Shipping on `main` | Assistant/tool orchestration | Standardized deployment, hierarchy-aware collaboration, shared ops plane |
+| **Codex** | Shipping on `main` | OpenAI Codex CLI workflows | Secure workspace abstraction, A2A delegation, org boundaries, shared control plane |
 | **Hermes 4** | Shipping on `main` | Hybrid reasoning, native tools, json_schema (NousResearch/hermes-agent) | Option B upstream hook, A2A bridge to OpenAI-compat API, multi-provider provider derivation |
-| **Gemini CLI** | Shipping on `main` | Google Gemini CLI continuity | Workspace lifecycle, A2A, hierarchy-aware collaboration, shared ops plane |
 | **OpenClaw** | Shipping on `main` | CLI-native runtime with its own session model | Workspace lifecycle, templates, activity logs, topology-aware collaboration |
 | **NemoClaw** | WIP on `feat/nemoclaw-t4-docker` | NVIDIA-oriented runtime path | Planned to join the same abstraction once merged; not yet part of `main` |
 
@@ -209,7 +205,7 @@ The result is not just “an agent that learns.” It is **an organization that
 ### Runtime
 
 - standalone workspace-template images that install `molecule-ai-workspace-runtime` from the Gitea package registry; thin AMI in production (us-east-2)
-- adapter-driven execution across **8 runtimes** (Claude Code, Hermes, Gemini CLI, LangGraph, DeepAgents, CrewAI, AutoGen, OpenClaw)
+- adapter-driven execution across **4 maintained runtimes** (Claude Code, Codex, Hermes, OpenClaw)
 - Agent Card registration
 - **Memory v2 backed by pgvector** — per-tenant plugin sidecar serving HMA namespaces with FTS + semantic recall
 - plugin-mounted shared rules/skills
@@ -245,7 +241,7 @@ The result is not just “an agent that learns.” It is **an organization that
 Molecule AI is especially strong when you need to run:
 
 - AI engineering teams with PM / Dev Lead / QA / Research / Ops roles
-- mixed runtime organizations where one team prefers LangGraph and another prefers Claude Code
+- mixed runtime organizations where one team prefers Hermes and another prefers Claude Code
 - long-lived agent organizations that need memory boundaries and reusable procedures
 - internal platforms that want to expose agent teams as structured infrastructure, not ad hoc scripts
 
@@ -260,7 +256,7 @@ Canvas (Next.js 15, warm-paper :3000)  <--HTTP / WS-->  Platform (Go 1.25 :8080)
          +------------------------- shows ------------------------> workspaces, teams, tasks, traces, events
 
 Workspace Runtime (Python ≥3.11, image with adapters)
-  - 8 adapters: LangGraph / DeepAgents / Claude Code / CrewAI / AutoGen / Hermes / Gemini CLI / OpenClaw
+  - 4 adapters: Claude Code / Codex / Hermes / OpenClaw
   - Agent Card + A2A server (typed-SSOT response path, RFC #2967)
   - heartbeat + activity + Memory v2 (pgvector semantic recall via per-tenant plugin sidecar)
   - skills + plugins + hot reload
@@ -328,7 +324,7 @@ Then open `http://localhost:3000`:
 
 ## Current Scope
 
-The current `main` branch ships the core platform, Canvas v4 (warm-paper themed), Memory v2 (pgvector semantic recall), the typed-SSOT A2A response path (RFC #2967), **eight production adapters** (Claude Code, Hermes, Gemini CLI, LangGraph, DeepAgents, CrewAI, AutoGen, OpenClaw), skill lifecycle, and operational surfaces.
+The current `main` branch ships the core platform, Canvas v4 (warm-paper themed), Memory v2 (pgvector semantic recall), the typed-SSOT A2A response path (RFC #2967), **four maintained production adapters** (Claude Code, Codex, Hermes, OpenClaw), skill lifecycle, and operational surfaces.
 
 The companion private repo [`molecule-controlplane`](https://git.moleculesai.app/molecule-ai/molecule-controlplane) provides the SaaS surface — multi-tenant orchestration on EC2 + Neon + Cloudflare Tunnels, KMS envelope encryption, WorkOS auth, Stripe billing, and a `tenant_resources` audit table with a 30-min reconciler.
 

README.zh-CN.md

@@ -52,7 +52,7 @@ Molecule AI 是目前最强的 AI Agent 组织治理方案之一，用来把 age
 它把过去分散在 demo、内部胶水代码和各类 framework 私有工具里的关键能力，收敛成一个产品：
 
 - 一套组织原生 control plane，管理团队、角色、层级和生命周期
-- 一套 runtime abstraction，让 **8 个** agent runtime —— LangGraph、DeepAgents、Claude Code、CrewAI、AutoGen、**Hermes**、**Gemini CLI**、OpenClaw —— 共用一套 workspace 契约
+- 一套 runtime abstraction，让 **4 个**维护中的 agent runtime —— Claude Code、Codex、**Hermes**、OpenClaw —— 共用一套 workspace 契约
 - 一套与组织边界对齐的 memory 模型，把 recall、sharing 和 skill evolution 放进同一体系（Memory v2 由 pgvector 支撑语义召回）
 - 一套面向线上 workspace 的运维面，统一完成观测、暂停、重启、检查和持续改进
 
@@ -74,7 +74,7 @@ Molecule AI 填的就是这个空白。
 
 ### 3. Runtime 选择不再是死路
 
-LangGraph、DeepAgents、Claude Code、CrewAI、AutoGen、Hermes、Gemini CLI、OpenClaw 都可以挂到同一个 workspace abstraction 下。团队可以统一治理方式，而不必统一到底层 runtime。
+Claude Code、Codex、Hermes、OpenClaw 都可以挂到同一个 workspace abstraction 下。团队可以统一治理方式，而不必统一到底层 runtime。
 
 ### 4. Memory 被当成基础设施来做
 
@@ -111,13 +111,9 @@ Molecule AI 并不是要替代下面这些 framework，而是把它们纳入更
 
 | Runtime / 架构 | 当前仓库状态 | 原生优势 | Molecule AI 额外补上的能力 |
 |---|---|---|---|
-| **LangGraph** | `main` 已支持 | 图控制强、工具调用成熟、Python 扩展性好 | Canvas orchestration、层级路由、A2A、memory scope、operational lifecycle |
-| **DeepAgents** | `main` 已支持 | 规划和任务拆解更强 | 同一套 workspace contract、团队拓扑、activity、restart 行为 |
 | **Claude Code** | `main` 已支持 | 真实编码工作流、CLI-native continuity | 安全 workspace 抽象、A2A delegation、组织边界、共享 control plane |
-| **CrewAI** | `main` 已支持 | 角色型 crew 模式清晰 | 持久 workspace 身份、统一策略、共享 Canvas 和 registry |
-| **AutoGen** | `main` 已支持 | assistant/tool orchestration | 统一部署、层级协作、共享运维平面 |
+| **Codex** | `main` 已支持 | OpenAI Codex CLI 工作流 | 安全 workspace 抽象、A2A delegation、组织边界、共享 control plane |
 | **Hermes 4** | `main` 已支持 | 混合推理、原生工具调用、json_schema 输出（NousResearch/hermes-agent） | Option B 上游 hook、A2A 桥接 OpenAI 兼容 API、多 provider 自动派生 |
-| **Gemini CLI** | `main` 已支持 | Google Gemini CLI 持续会话 | workspace 生命周期、A2A、层级感知协作、共享运维平面 |
 | **OpenClaw** | `main` 已支持 | CLI-native runtime，自有 session 模型 | workspace 生命周期、templates、activity logs、拓扑感知协作 |
 | **NemoClaw** | `feat/nemoclaw-t4-docker` 分支 WIP | NVIDIA 方向 runtime 路线 | 计划并入同一抽象层，但当前还不是 `main` 已合并能力 |
 
@@ -208,7 +204,7 @@ Molecule AI 并不是要替代下面这些 framework，而是把它们纳入更
 ### Runtime
 
 - 统一 `workspace/` 镜像；生产环境采用 thin AMI（us-east-2）
-- adapter 驱动执行，覆盖 **8 个 runtime**（Claude Code、Hermes、Gemini CLI、LangGraph、DeepAgents、CrewAI、AutoGen、OpenClaw）
+- adapter 驱动执行，覆盖 **4 个维护中的 runtime**（Claude Code、Codex、Hermes、OpenClaw）
 - Agent Card 注册
 - **Memory v2 由 pgvector 支撑** —— 每个 tenant 一个 plugin sidecar，承载 HMA namespace、FTS 与语义召回
 - plugin 挂载共享 rules/skills
@@ -259,7 +255,7 @@ Canvas (Next.js 15, warm-paper :3000)  <--HTTP / WS-->  Platform (Go 1.25 :8080)
          +------------------------- 展示 ------------------------> workspaces, teams, tasks, traces, events
 
 Workspace Runtime (Python ≥3.11，含 adapter 集合的镜像)
-  - 8 个 adapter: LangGraph / DeepAgents / Claude Code / CrewAI / AutoGen / Hermes / Gemini CLI / OpenClaw
+  - 4 个 adapter: Claude Code / Codex / Hermes / OpenClaw
   - Agent Card + A2A server（typed-SSOT 响应路径，RFC #2967）
   - heartbeat + activity + Memory v2（pgvector 语义召回，per-tenant plugin sidecar）
   - skills + plugins + hot reload
@@ -321,7 +317,7 @@ npm run dev
 
 ## 当前范围说明
 
-当前 `main` 已经包含核心平台、Canvas v4（warm-paper 主题）、Memory v2（pgvector 语义召回）、typed-SSOT A2A 响应路径（RFC #2967）、**8 个正式 adapter**（Claude Code、Hermes、Gemini CLI、LangGraph、DeepAgents、CrewAI、AutoGen、OpenClaw）、skill lifecycle，以及主要运维面。
+当前 `main` 已经包含核心平台、Canvas v4（warm-paper 主题）、Memory v2（pgvector 语义召回）、typed-SSOT A2A 响应路径（RFC #2967）、**4 个维护中的正式 adapter**（Claude Code、Codex、Hermes、OpenClaw）、skill lifecycle，以及主要运维面。
 
 配套的私有仓库 [`molecule-controlplane`](https://git.moleculesai.app/molecule-ai/molecule-controlplane) 提供 SaaS 层 —— 多租户编排（EC2 + Neon + Cloudflare Tunnels）、KMS 信封加密、WorkOS 鉴权、Stripe 计费，以及 `tenant_resources` 审计表加 30 分钟 reconciler。
 

canvas/src/components/tabs/ConfigTab.tsx

@@ -301,16 +301,13 @@ export function deriveProvidersFromModels(models: ModelSpec[]): string[] {
 // config.yaml` on the container is a separate runtime-internal file,
 // not this one.
 const RUNTIMES_WITH_OWN_CONFIG = new Set<string>(["external", "kimi", "kimi-cli", "openclaw"]);
+const SUPPORTED_RUNTIME_VALUES = new Set(["claude-code", "codex", "openclaw", "hermes"]);
 
 const FALLBACK_RUNTIME_OPTIONS: RuntimeOption[] = [
-  { value: "", label: "LangGraph (default)", models: [], providers: [] },
   { value: "claude-code", label: "Claude Code", models: [], providers: [] },
-  { value: "crewai", label: "CrewAI", models: [], providers: [] },
-  { value: "autogen", label: "AutoGen", models: [], providers: [] },
-  { value: "deepagents", label: "DeepAgents", models: [], providers: [] },
+  { value: "codex", label: "Codex", models: [], providers: [] },
   { value: "openclaw", label: "OpenClaw", models: [], providers: [] },
   { value: "hermes", label: "Hermes", models: [], providers: [] },
-  { value: "gemini-cli", label: "Gemini CLI", models: [], providers: [] },
 ];
 
 export function ConfigTab({ workspaceId }: Props) {
@@ -499,10 +496,9 @@ export function ConfigTab({ workspaceId }: Props) {
       .then((rows) => {
         if (cancelled || !Array.isArray(rows)) return;
         const byRuntime = new Map<string, RuntimeOption>();
-        byRuntime.set("", { value: "", label: "LangGraph (default)", models: [], providers: [] });
         for (const r of rows) {
           const v = (r.runtime || "").trim();
-          if (!v || v === "langgraph") continue;
+          if (!SUPPORTED_RUNTIME_VALUES.has(v)) continue;
           // Last template wins if two templates share a runtime — rare, and the
           // one with the richer models list is probably newer.
           const existing = byRuntime.get(v);
@@ -512,7 +508,7 @@ export function ConfigTab({ workspaceId }: Props) {
             byRuntime.set(v, { value: v, label: r.name || v, models, providers });
           }
         }
-        if (byRuntime.size > 1) setRuntimeOptions(Array.from(byRuntime.values()));
+        if (byRuntime.size > 0) setRuntimeOptions(Array.from(byRuntime.values()));
       })
       .catch(() => { /* keep fallback */ });
     return () => { cancelled = true; };

canvas/src/lib/__tests__/preflight-resolveRuntime.test.ts

@@ -12,47 +12,35 @@ import { resolveRuntime } from "../deploy-preflight";
 
 describe("resolveRuntime", () => {
   describe("explicit runtime-map entries", () => {
-    it('maps "langgraph" to "langgraph"', () => {
-      expect(resolveRuntime("langgraph")).toBe("langgraph");
-    });
-
     it('maps "claude-code-default" to "claude-code"', () => {
       expect(resolveRuntime("claude-code-default")).toBe("claude-code");
     });
 
+    it('maps "codex" to "codex"', () => {
+      expect(resolveRuntime("codex")).toBe("codex");
+    });
+
+    it('maps "hermes" to "hermes"', () => {
+      expect(resolveRuntime("hermes")).toBe("hermes");
+    });
+
     it('maps "openclaw" to "openclaw"', () => {
       expect(resolveRuntime("openclaw")).toBe("openclaw");
     });
-
-    it('maps "deepagents" to "deepagents"', () => {
-      expect(resolveRuntime("deepagents")).toBe("deepagents");
-    });
-
-    it('maps "crewai" to "crewai"', () => {
-      expect(resolveRuntime("crewai")).toBe("crewai");
-    });
-
-    it('maps "autogen" to "autogen"', () => {
-      expect(resolveRuntime("autogen")).toBe("autogen");
-    });
   });
 
   describe("identity fallback for modern template ids", () => {
-    it("returns the id unchanged when not in the map", () => {
-      expect(resolveRuntime("hermes")).toBe("hermes");
-    });
-
     it("strips trailing -default suffix as fallback", () => {
       expect(resolveRuntime("hermes-default")).toBe("hermes");
     });
 
     it("strips -default only when it is the suffix", () => {
       // "default-something" should NOT strip
-      expect(resolveRuntime("default-langgraph")).toBe("default-langgraph");
+      expect(resolveRuntime("default-custom")).toBe("default-custom");
     });
 
     it("returns the id unchanged when id has no -default suffix", () => {
-      expect(resolveRuntime("gemini-cli")).toBe("gemini-cli");
+      expect(resolveRuntime("custom-runtime")).toBe("custom-runtime");
     });
 
     it("handles custom template ids from community templates", () => {

canvas/src/lib/__tests__/runtime-names.test.ts

@@ -13,11 +13,9 @@ import { runtimeDisplayName } from "../runtime-names";
 describe("runtimeDisplayName", () => {
   it.each([
     ["claude-code", "Claude Code"],
-    ["langgraph", "LangGraph"],
-    ["deepagents", "DeepAgents"],
+    ["codex", "Codex"],
+    ["hermes", "Hermes"],
     ["openclaw", "OpenClaw"],
-    ["crewai", "CrewAI"],
-    ["autogen", "AutoGen"],
   ])("known runtime %q maps to %q", (input, expected) => {
     expect(runtimeDisplayName(input)).toBe(expected);
   });
@@ -25,7 +23,6 @@ describe("runtimeDisplayName", () => {
   it("unknown runtime falls back to the input string verbatim", () => {
     // A future runtime not yet in the lookup map should render with
     // its own id — better than a generic placeholder for ops debugging.
-    expect(runtimeDisplayName("hermes")).toBe("hermes");
     expect(runtimeDisplayName("custom-runtime-9000")).toBe(
       "custom-runtime-9000",
     );
@@ -43,6 +40,6 @@ describe("runtimeDisplayName", () => {
     // the input "for safety" doesn't silently change behavior — the
     // upstream slug is already normalized lowercase.
     expect(runtimeDisplayName("Claude-Code")).toBe("Claude-Code");
-    expect(runtimeDisplayName("LANGGRAPH")).toBe("LANGGRAPH");
+    expect(runtimeDisplayName("CODEX")).toBe("CODEX");
   });
 });

canvas/src/lib/deploy-preflight.ts

@@ -63,12 +63,10 @@ export interface Template extends TemplateLike {
  *  id needs a non-identity mapping. */
 export function resolveRuntime(templateId: string): string {
   const runtimeMap: Record<string, string> = {
-    langgraph: "langgraph",
     "claude-code-default": "claude-code",
+    codex: "codex",
+    hermes: "hermes",
     openclaw: "openclaw",
-    deepagents: "deepagents",
-    crewai: "crewai",
-    autogen: "autogen",
   };
   return runtimeMap[templateId] ?? templateId.replace(/-default$/, "");
 }

canvas/src/lib/runtime-names.ts

@@ -4,11 +4,9 @@
 
 const RUNTIME_NAMES: Record<string, string> = {
   "claude-code": "Claude Code",
-  langgraph: "LangGraph",
-  deepagents: "DeepAgents",
+  codex: "Codex",
+  hermes: "Hermes",
   openclaw: "OpenClaw",
-  crewai: "CrewAI",
-  autogen: "AutoGen",
   kimi: "Kimi",
   "kimi-cli": "Kimi CLI",
 };

canvas/src/lib/runtimeProfiles.ts

@@ -49,7 +49,7 @@ export interface RuntimeProfile {
 }
 
 /** The floor every runtime inherits unless it overrides. Calibrated for
- *  docker-local fast runtimes (claude-code, langgraph, crewai) where cold
+ *  docker-local fast runtimes (claude-code, codex, openclaw) where cold
  *  boot is 30-90s. */
 export const DEFAULT_RUNTIME_PROFILE: Required<
   Pick<RuntimeProfile, "provisionTimeoutMs">

docs/agent-runtime/cli-runtime.md

@@ -2,14 +2,14 @@
 
 ## Overview
 
-The workspace runtime uses a **pluggable adapter architecture** — each agent infrastructure (Claude Code, OpenClaw, LangGraph, CrewAI, AutoGen, etc.) has its own adapter that bridges the A2A protocol to the infra's native interface.
+The workspace runtime uses a **pluggable adapter architecture** — each maintained agent infrastructure (Claude Code, Codex, Hermes, OpenClaw) has its own adapter that bridges the A2A protocol to the infra's native interface.
 
 Adapters live in `workspace/adapters/<runtime>/` and are auto-discovered at startup. Each adapter implements `BaseAdapter` (from `adapters/base.py`) with `setup()` and `create_executor()` methods.
 
 The runtime is selected via `config.yaml`:
 
 ```yaml
-runtime: claude-code    # or: langgraph, openclaw, deepagents, crewai, autogen
+runtime: claude-code    # or: codex, hermes, openclaw
 runtime_config:
   model: sonnet
   auth_token_file: .auth-token
@@ -18,7 +18,7 @@ runtime_config:
 
 ## How It Works
 
-The unified `workspace-template` Docker image includes both Python (LangGraph) and Node.js (CLI runtimes). At startup, `main.py` checks the `runtime` field in `config.yaml`, discovers the matching adapter in `adapters/<runtime>/`, calls `adapter.setup(config)` then `adapter.create_executor(config)` to get an `AgentExecutor` that handles A2A requests.
+The unified runtime checks the `runtime` field in `config.yaml`, discovers the matching adapter, calls `adapter.setup(config)` then `adapter.create_executor(config)` to get an `AgentExecutor` that handles A2A requests.
 
 ```
 A2A request arrives
@@ -28,7 +28,7 @@ AgentExecutor.execute(context, event_queue)
       |  - extracts user message from A2A parts
       |  - extracts conversation history from params.metadata.history
       |  - sets current_task on heartbeat (shows on canvas card)
-      |  - invokes the runtime (LangGraph graph, CLI subprocess, etc.)
+      |  - invokes the runtime adapter
       v
 Response → A2A event queue → JSON-RPC response
 ```
@@ -37,9 +37,9 @@ Response → A2A event queue → JSON-RPC response
 
 Chat sessions in the Canvas UI send prior messages (up to 20) via `params.metadata.history` in each A2A `message/send` request. Executors extract this history:
 
-- **LangGraph/DeepAgents**: Prepends history as `("human", text)` / `("ai", text)` tuples to the LangGraph message list
-- **CrewAI/AutoGen**: Prepends history as a text prefix in the task description (`"Conversation so far:\n..."`)
 - **Claude Code**: Uses `--resume <session_id>` for native session continuity (history not needed)
+- **Codex**: Uses the Codex runtime's native session state
+- **Hermes**: Uses Hermes' agent runtime session handling
 - **OpenClaw**: Uses `--session-id` for native session continuity
 
 ### Current Task Reporting
@@ -48,10 +48,6 @@ All executors update the workspace's `current_task` via the heartbeat during exe
 
 ## Built-in Adapters
 
-### LangGraph (`runtime: langgraph`) — Default
-
-Full Python agent with LangGraph ReAct pattern. Supports skills, tools, plugins, peer coordination, and team routing.
-
 ### Claude Code (`runtime: claude-code`)
 
 ```yaml
@@ -71,35 +67,18 @@ The SDK uses the same Claude Code engine under the hood — plugins, CLAUDE.md d
 
 **Important:** Claude Code refuses to run as root with `--dangerously-skip-permissions`. The Dockerfile creates a non-root `agent` user.
 
-### CrewAI (`runtime: crewai`)
-
-Role-based multi-agent framework. Creates a CrewAI Agent + Task + Crew per request with A2A delegation tools (`delegate_to_peer`, `list_available_peers`).
+### Codex (`runtime: codex`)
 
 ```yaml
-runtime: crewai
-model: openrouter:google/gemini-2.5-flash
+runtime: codex
+model: openai/gpt-5.3-codex
 ```
 
-**Auth:** Uses `OPENROUTER_API_KEY` or `OPENAI_API_KEY` env var.
-
-### AutoGen (`runtime: autogen`)
-
-Microsoft AutoGen AssistantAgent with tool use. Creates an `AssistantAgent` per request with A2A delegation tools.
+### Hermes (`runtime: hermes`)
 
 ```yaml
-runtime: autogen
-model: openai:gpt-4.1-mini
-```
-
-**Auth:** Uses `OPENAI_API_KEY` env var.
-
-### DeepAgents (`runtime: deepagents`)
-
-LangGraph-based agent with deep planning capabilities. Uses the same `LangGraphA2AExecutor` as the default runtime but with a specialized agent setup including delegation, memory, and search tools.
-
-```yaml
-runtime: deepagents
-model: openrouter:google/gemini-2.5-flash
+runtime: hermes
+model: openai/gpt-4o
 ```
 
 ### OpenClaw (`runtime: openclaw`)

docs/agent-runtime/workspace-runtime.md

@@ -4,13 +4,11 @@ The `workspace/` directory is Molecule AI's unified runtime image. Every provisi
 
 ## Runtime Matrix In Current `main`
 
-Current `main` ships six adapters:
+Current `main` ships four maintained adapters:
 
-- `langgraph`
-- `deepagents`
 - `claude-code`
-- `crewai`
-- `autogen`
+- `codex`
+- `hermes`
 - `openclaw`
 
 This is the merged runtime surface today. Branch-level experiments such as NemoClaw are separate and should be treated as roadmap/WIP, not merged support.

docs/architecture/molecule-technical-doc.md

@@ -511,7 +511,7 @@ description: ""
 version: "1.0.0"
 tier: 2                                    # 1=sandboxed, 2=standard, 3=privileged, 4=full-host
 model: "anthropic:claude-sonnet-4-6"       # provider:model syntax
-runtime: "langgraph"                       # langgraph | deepagents | claude-code | crewai | autogen | openclaw
+runtime: "claude-code"                     # claude-code | codex | hermes | openclaw
 runtime_config:                            # Runtime-specific settings
   command: "claude"                        # For CLI runtimes
   args: []
@@ -565,15 +565,13 @@ compliance:
   max_task_duration_seconds: 300
 ```
 
-### Six Runtime Adapters
+### Four Runtime Adapters
 
 | Adapter | Core Strength | Image Tag |
 |---------|--------------|-----------|
-| **LangGraph** | Graph-based state machine, tool use, streaming | `workspace-template:langgraph` |
-| **DeepAgents** | Deep planning, multi-step task decomposition | `workspace-template:deepagents` |
 | **Claude Code** | Native coding workflows, CLI continuity, OAuth auth | `workspace-template:claude-code` |
-| **CrewAI** | Role-based crews, structured task orchestration | `workspace-template:crewai` |
-| **AutoGen** | Multi-agent conversations, explicit strategies | `workspace-template:autogen` |
+| **Codex** | OpenAI Codex coding workflows | `workspace-template:codex` |
+| **Hermes** | Hermes agent runtime | `workspace-template:hermes` |
 | **OpenClaw** | CLI-native runtime, own session model | `workspace-template:openclaw` |
 
 **Branch-level WIP**: NemoClaw (NVIDIA T4 + Docker socket) on `feat/nemoclaw-t4-docker`.

docs/architecture/overview.md

@@ -44,7 +44,7 @@ All three call `onWorkspaceOffline`, which broadcasts `WORKSPACE_OFFLINE` and ca
 
 ### Template Resolution (Workspace Create)
 
-Runtime detection happens **before** the DB insert: if `payload.Runtime` is empty and a template is specified, the handler reads `runtime:` from `configsDir/template/config.yaml` first. If still empty, it defaults to `"langgraph"`. This ensures the correct runtime (e.g. `claude-code`) is persisted in the DB and used for container image selection.
+Runtime detection happens **before** the DB insert: if `payload.Runtime` is empty and a template is specified, the handler reads `runtime:` from `configsDir/template/config.yaml` first. If still empty, it defaults to `"claude-code"`. This ensures the correct runtime is persisted in the DB and used for container image selection.
 
 When the requested template does not exist, the Create handler falls back in order:
 

docs/index.md

@@ -24,7 +24,7 @@ features:
     details: Build agent organizations as nested workspaces on a live React Flow canvas with drag-to-nest hierarchy, template deployment, bundles, and real-time updates.
     icon: "🗺️"
   - title: Runtime Compatibility
-    details: Current main ships adapters for LangGraph, DeepAgents, Claude Code, CrewAI, AutoGen, and OpenClaw under one workspace contract and A2A surface.
+    details: Current main ships adapters for Claude Code, Codex, Hermes, and OpenClaw under one workspace contract and A2A surface.
     icon: "⚙️"
   - title: Hierarchical Memory
     details: HMA-style LOCAL, TEAM, and GLOBAL scopes backed by the v2 memory plugin (per-tenant pgvector sidecar with FTS + semantic recall).
@@ -49,13 +49,13 @@ features:
 |---|---|
 | **Canvas** | Empty-state deployment, onboarding guide, 10-tab side panel, template palette, bundle import/export, drag-to-nest teams, search, activity and trace views |
 | **Platform** | Workspace CRUD, registry, A2A proxy, team expansion, approvals, secrets, global secrets, memory APIs, files API, terminal, viewport persistence, WebSocket fanout |
-| **Runtime** | One workspace image with six shipping adapters on `main`: LangGraph, DeepAgents, Claude Code, CrewAI, AutoGen, OpenClaw |
+| **Runtime** | One workspace image with four shipping adapters on `main`: Claude Code, Codex, Hermes, OpenClaw |
 | **Memory** | v2 plugin (pgvector + FTS) serving scoped agent memories under per-workspace namespaces; key/value workspace memory; session-search recall |
 | **Skills** | Local skill packages, plugin-mounted shared skills/rules, audit/install/publish CLI helpers, hot reload |
 
 ## Compatibility Note
 
-`main` currently ships six runtime adapters. `NemoClaw` appears in branch-level work (`feat/nemoclaw-t4-docker`) and is not documented here as merged `main` functionality.
+`main` currently ships four runtime adapters: Claude Code, Codex, Hermes, and OpenClaw. `NemoClaw` appears in branch-level work (`feat/nemoclaw-t4-docker`) and is not documented here as merged `main` functionality.
 
 ## Recommended Reading
 

docs/quickstart.md

@@ -238,7 +238,7 @@ No inbound firewall rules needed — the agent initiates the outbound WebSocket
 ## What To Try Next
 
 - **Expand to a team:** right-click a workspace and choose `Expand to Team`.
-- **Switch runtime:** use `Config -> Runtime` to move between LangGraph, DeepAgents, Claude Code, CrewAI, AutoGen, and OpenClaw.
+- **Switch runtime:** use `Config -> Runtime` to move between Claude Code, Codex, Hermes, and OpenClaw.
 - **Inspect operations:** check `Activity`, `Traces`, `Events`, and `Terminal`.
 - **Use global keys:** configure one provider once in `Secrets & API Keys -> Global`.
 - **Import a template:** use the template palette or `POST /templates/import`.
@@ -268,7 +268,7 @@ Browser  -->  Canvas (Next.js :3000)
                                 |
                                 v
                        Provisioned workspaces
-                     (LangGraph / Claude Code / CrewAI / AutoGen / etc.)
+                     (Claude Code / Codex / Hermes / OpenClaw)
 ```
 
 For the full system model, see [Architecture](./architecture/architecture.md).

scripts/build-images.sh

@@ -8,7 +8,7 @@ cd "$(dirname "$0")/../workspace-template"
 echo "=== Building base image ==="
 docker build -t workspace-template:base -t workspace-template:latest .
 
-for adapter in langgraph claude_code openclaw deepagents crewai autogen; do
+for adapter in claude_code codex hermes openclaw; do
   DOCKERFILE="adapters/${adapter}/Dockerfile"
   if [ ! -f "$DOCKERFILE" ]; then
     echo "Skipping $adapter (no Dockerfile)"

scripts/refresh-workspace-images.sh

@@ -32,7 +32,7 @@ log()  { echo -e "${GREEN}[refresh]${NC} $1" >&2; }
 warn() { echo -e "${YELLOW}[refresh]${NC} $1" >&2; }
 err()  { echo -e "${RED}[refresh]${NC} $1" >&2; }
 
-ALL_RUNTIMES=(claude-code langgraph crewai autogen deepagents hermes gemini-cli openclaw)
+ALL_RUNTIMES=(claude-code codex hermes openclaw)
 RUNTIMES=("${ALL_RUNTIMES[@]}")
 RECREATE=true
 

scripts/test-all-adapters.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# E2E test: All 6 adapters — create one agent per runtime, test A2A between all
+# E2E test: all maintained adapters — create one agent per runtime, test A2A
 set -euo pipefail
 
 PLATFORM="${1:-http://localhost:8080}"
@@ -52,12 +52,12 @@ a2a_send() {
 }
 
 echo "============================================"
-echo "  All-Adapters E2E Test (6 runtimes)"
+echo "  All-Adapters E2E Test (4 runtimes)"
 echo "============================================"
 echo ""
 
 # --- Create workspaces ---
-echo "--- Step 1: Create 6 workspaces ---"
+echo "--- Step 1: Create 4 workspaces ---"
 
 R=$(curl -s -X POST "$PLATFORM/workspaces" -H 'Content-Type: application/json' \
   -d '{"name":"Alice-Claude","role":"claude-code test","tier":2,"template":"claude-code-default"}')
@@ -65,9 +65,9 @@ ALICE=$(echo "$R" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'
 check "Create Alice (claude-code)" "provisioning" "$R"
 
 R=$(curl -s -X POST "$PLATFORM/workspaces" -H 'Content-Type: application/json' \
-  -d '{"name":"Bob-LangGraph","role":"langgraph test","tier":2,"template":"langgraph"}')
+  -d '{"name":"Bob-Codex","role":"codex test","tier":2,"template":"codex"}')
 BOB=$(echo "$R" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
-check "Create Bob (langgraph)" "provisioning" "$R"
+check "Create Bob (codex)" "provisioning" "$R"
 
 R=$(curl -s -X POST "$PLATFORM/workspaces" -H 'Content-Type: application/json' \
   -d '{"name":"Carol-OpenClaw","role":"openclaw test","tier":2,"template":"openclaw"}')
@@ -75,29 +75,19 @@ CAROL=$(echo "$R" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'
 check "Create Carol (openclaw)" "provisioning" "$R"
 
 R=$(curl -s -X POST "$PLATFORM/workspaces" -H 'Content-Type: application/json' \
-  -d '{"name":"Dave-DeepAgents","role":"deepagents test","tier":2,"template":"deepagents"}')
+  -d '{"name":"Dave-Hermes","role":"hermes test","tier":2,"template":"hermes"}')
 DAVE=$(echo "$R" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
-check "Create Dave (deepagents)" "provisioning" "$R"
-
-R=$(curl -s -X POST "$PLATFORM/workspaces" -H 'Content-Type: application/json' \
-  -d '{"name":"Eve-CrewAI","role":"crewai test","tier":2,"template":"crewai"}')
-EVE=$(echo "$R" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
-check "Create Eve (crewai)" "provisioning" "$R"
-
-R=$(curl -s -X POST "$PLATFORM/workspaces" -H 'Content-Type: application/json' \
-  -d '{"name":"Frank-AutoGen","role":"autogen test","tier":2,"template":"autogen"}')
-FRANK=$(echo "$R" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
-check "Create Frank (autogen)" "provisioning" "$R"
+check "Create Dave (hermes)" "provisioning" "$R"
 
 # --- Set API keys (skip Claude which uses OAuth) ---
 echo ""
 echo "--- Step 2: Set API keys ---"
-for ID in $BOB $CAROL $DAVE $EVE $FRANK; do
+for ID in $BOB $CAROL $DAVE; do
   curl -s -X POST "$PLATFORM/workspaces/$ID/secrets" \
     -H 'Content-Type: application/json' \
     -d "{\"key\":\"OPENAI_API_KEY\",\"value\":\"$OPENAI_KEY\"}" > /dev/null
 done
-echo "Set OPENAI_API_KEY on 5 agents"
+echo "Set OPENAI_API_KEY on 3 agents"
 
 # Auto-restart happens automatically when secrets are set
 echo "Secrets trigger auto-restart — waiting for agents to come back..."
@@ -105,13 +95,11 @@ sleep 15
 
 # --- Wait for all online ---
 echo ""
-echo "--- Step 3: Wait for agents (OpenClaw ~3min, CrewAI/AutoGen/DeepAgents ~2min) ---"
+echo "--- Step 3: Wait for agents (OpenClaw ~3min, Hermes may take longer) ---"
 
 wait_online "$ALICE" "Alice-Claude" 20 && check "Alice online" "ok" "ok" || check "Alice online" "online" "timeout"
-wait_online "$BOB" "Bob-LangGraph" 60 && check "Bob online" "ok" "ok" || check "Bob online" "online" "timeout"
-wait_online "$DAVE" "Dave-DeepAgents" 120 && check "Dave online" "ok" "ok" || check "Dave online" "online" "timeout"
-wait_online "$EVE" "Eve-CrewAI" 120 && check "Eve online" "ok" "ok" || check "Eve online" "online" "timeout"
-wait_online "$FRANK" "Frank-AutoGen" 120 && check "Frank online" "ok" "ok" || check "Frank online" "online" "timeout"
+wait_online "$BOB" "Bob-Codex" 60 && check "Bob online" "ok" "ok" || check "Bob online" "online" "timeout"
+wait_online "$DAVE" "Dave-Hermes" 180 && check "Dave online" "ok" "ok" || check "Dave online" "online" "timeout"
 wait_online "$CAROL" "Carol-OpenClaw" 360 && check "Carol online" "ok" "ok" || check "Carol online" "online" "timeout"
 
 # --- Test A2A messages ---
@@ -123,7 +111,7 @@ RESP=$(a2a_send "$ALICE" "say hello in one word")
 echo "    -> $RESP"
 check "Alice responds" "hello" "$RESP"
 
-echo "  Talking to Bob (LangGraph)..."
+echo "  Talking to Bob (Codex)..."
 RESP=$(a2a_send "$BOB" "say hello in one word")
 echo "    -> $RESP"
 check "Bob responds" "hello" "$RESP"
@@ -133,21 +121,11 @@ RESP=$(a2a_send "$CAROL" "say hello in one word")
 echo "    -> $RESP"
 check "Carol responds" "hello" "$RESP"
 
-echo "  Talking to Dave (DeepAgents)..."
+echo "  Talking to Dave (Hermes)..."
 RESP=$(a2a_send "$DAVE" "say hello in one word")
 echo "    -> $RESP"
 check "Dave responds" "hello" "$RESP"
 
-echo "  Talking to Eve (CrewAI)..."
-RESP=$(a2a_send "$EVE" "say hello in one word")
-echo "    -> $RESP"
-check "Eve responds" "hello" "$RESP"
-
-echo "  Talking to Frank (AutoGen)..."
-RESP=$(a2a_send "$FRANK" "say hello in one word")
-echo "    -> $RESP"
-check "Frank responds" "hello" "$RESP"
-
 # --- Peer discovery ---
 echo ""
 echo "--- Step 5: Peer discovery ---"
@@ -157,7 +135,7 @@ peers = json.load(sys.stdin)
 print(f'{len(peers)} peers: {\" \".join(p.get(\"name\",\"\") for p in peers)}')
 " 2>/dev/null)
 echo "  Alice sees: $R"
-check "Alice sees 5 peers" "5 peers" "$R"
+check "Alice sees 3 peers" "3 peers" "$R"
 
 # --- Isolation ---
 echo ""
@@ -168,7 +146,7 @@ check "No ws-* dirs on host" "0" "$HOST_WS"
 # --- Cleanup ---
 echo ""
 echo "--- Step 7: Cleanup ---"
-for ID in $ALICE $BOB $CAROL $DAVE $EVE $FRANK; do
+for ID in $ALICE $BOB $CAROL $DAVE; do
   curl -s -X DELETE "$PLATFORM/workspaces/$ID" > /dev/null 2>&1
 done
 check "Cleanup" "ok" "ok"

tests/e2e/lib/model_slug.sh

@@ -10,15 +10,6 @@
 #                                    "gpt-4o" falls through to Anthropic
 #                                    default + 401, see PR #1714.)
 #
-#   langgraph   → "openai:gpt-4o"  (colon-form: langchain init_chat_model
-#                                    requires "<provider>:<model>".
-#                                    Slash-form was misinterpreted as
-#                                    OpenRouter routing → fell through
-#                                    without auth, surfaced 2026-05-03
-#                                    after the a2a-sdk v1 contract bugs
-#                                    PR #2558+#2563+#2567 cleared the
-#                                    masking layers.)
-#
 #   claude-code → auth-aware:
 #                  E2E_MINIMAX_API_KEY    → "MiniMax-M2"
 #                  E2E_ANTHROPIC_API_KEY  → "claude-sonnet-4-6"
@@ -51,7 +42,6 @@ pick_model_slug() {
   fi
   case "$runtime" in
     hermes)      printf 'openai/gpt-4o' ;;
-    langgraph)   printf 'openai:gpt-4o' ;;
     claude-code)
       if [ -n "${E2E_MINIMAX_API_KEY:-}" ]; then
         printf 'MiniMax-M2'

tests/e2e/test_chat_upload_e2e.sh

@@ -15,7 +15,7 @@
 # Required env:
 #   BASE                   default http://localhost:8080
 #                          override to https://<id>.<tenant>.staging...
-#   WORKSPACE_RUNTIME      default langgraph (any internal runtime)
+#   WORKSPACE_RUNTIME      default claude-code (any maintained internal runtime)
 #
 # Exit codes:
 #   0  upload + read-back round-trip succeeded
@@ -26,7 +26,7 @@
 set -uo pipefail
 
 BASE="${BASE:-http://localhost:8080}"
-RUNTIME="${WORKSPACE_RUNTIME:-langgraph}"
+RUNTIME="${WORKSPACE_RUNTIME:-claude-code}"
 
 PARENT=""
 PARENT_TOK=""
@@ -49,7 +49,7 @@ trap cleanup EXIT INT TERM
 echo "[1/5] POST /workspaces (runtime=$RUNTIME)..."
 P_RESP=$(curl -sS -X POST "$BASE/workspaces" \
     -H "Content-Type: application/json" \
-    -d "{\"name\":\"e2e-chat-upload\",\"runtime\":\"$RUNTIME\",\"tier\":2}")
+    -d "{\"name\":\"e2e-chat-upload\",\"runtime\":\"$RUNTIME\",\"tier\":2,\"model\":\"sonnet\"}")
 PARENT=$(echo "$P_RESP" | python3 -c "import sys,json; print(json.load(sys.stdin).get('id',''))" 2>/dev/null)
 [ -n "$PARENT" ] || { echo "  ✗ workspace create failed: $P_RESP"; exit 1; }
 echo "  ✓ workspace=$PARENT"

tests/e2e/test_comprehensive_e2e.sh

@@ -137,14 +137,14 @@ check "Create claude-code workspace" '"status":"provisioning"' "$R"
 RT_CC_ID=$(echo "$R" | jq_extract "['id']")
 
 R=$(curl -s -X POST "$BASE/workspaces" -H "Content-Type: application/json" \
-  -d '{"name":"RT LangGraph","role":"Test","tier":2,"runtime":"langgraph"}')
-check "Create langgraph workspace" '"status":"provisioning"' "$R"
-RT_LG_ID=$(echo "$R" | jq_extract "['id']")
+  -d '{"name":"RT Codex","role":"Test","tier":2,"runtime":"codex"}')
+check "Create codex workspace" '"status":"provisioning"' "$R"
+RT_CX_ID=$(echo "$R" | jq_extract "['id']")
 
 R=$(curl -s -X POST "$BASE/workspaces" -H "Content-Type: application/json" \
-  -d '{"name":"RT CrewAI","role":"Test","tier":2,"runtime":"crewai"}')
-check "Create crewai workspace" '"status":"provisioning"' "$R"
-RT_CR_ID=$(echo "$R" | jq_extract "['id']")
+  -d '{"name":"RT Hermes","role":"Test","tier":2,"runtime":"hermes"}')
+check "Create hermes workspace" '"status":"provisioning"' "$R"
+RT_HM_ID=$(echo "$R" | jq_extract "['id']")
 
 # Wait for containers to start (poll up to 30s for first one to appear)
 if command -v docker &>/dev/null; then
@@ -174,8 +174,8 @@ if command -v docker &>/dev/null; then
   }
 
   _check_image "$RT_CC_ID" "claude-code" "claude-code uses claude-code image"
-  _check_image "$RT_LG_ID" "langgraph" "langgraph uses langgraph image"
-  _check_image "$RT_CR_ID" "crewai" "crewai uses crewai image"
+  _check_image "$RT_CX_ID" "codex" "codex uses codex image"
+  _check_image "$RT_HM_ID" "hermes" "hermes uses hermes image"
 else
   echo "  SKIP: Docker not available — cannot verify container images"
   SKIP=$((SKIP + 3))
@@ -183,7 +183,7 @@ fi
 
 # Verify runtime in agent card after registration
 sleep 5
-for rt_id in $RT_CC_ID $RT_LG_ID $RT_CR_ID; do
+for rt_id in $RT_CC_ID $RT_CX_ID $RT_HM_ID; do
   # Register so we can check agent card
   curl -s -X POST "$BASE/registry/register" -H "Content-Type: application/json" \
     -d "{\"id\":\"$rt_id\",\"url\":\"http://localhost:19999\",\"agent_card\":{\"name\":\"Test\",\"skills\":[]}}" > /dev/null 2>&1
@@ -204,20 +204,20 @@ fi
 
 # Verify runtime change persists on restart (if provisioner supports ExecRead)
 # Write a new runtime to config, restart, check image changes
-R=$(curl -s -X PUT "$BASE/workspaces/$RT_LG_ID/files/config.yaml" \
+R=$(curl -s -X PUT "$BASE/workspaces/$RT_CX_ID/files/config.yaml" \
   -H "Content-Type: application/json" \
-  -d '{"content":"name: RT LangGraph\nruntime: deepagents\nmodel: openai:gpt-4.1-mini\ntier: 2\n"}')
+  -d '{"content":"name: RT Codex\nruntime: openclaw\nmodel: openai:gpt-4.1-mini\ntier: 2\n"}')
 if echo "$R" | grep -qF "saved"; then
-  curl -s -X POST "$BASE/workspaces/$RT_LG_ID/restart" > /dev/null 2>&1
+  curl -s -X POST "$BASE/workspaces/$RT_CX_ID/restart" > /dev/null 2>&1
   # Poll up to 30s for the new container image to appear (restart can take a while)
   if command -v docker &>/dev/null; then
-    short_id="${RT_LG_ID:0:12}"
+    short_id="${RT_CX_ID:0:12}"
     for _ in 1 2 3 4 5 6; do
       sleep 5
       actual=$(docker inspect "ws-${short_id}" --format '{{.Config.Image}}' 2>/dev/null || echo "")
-      if echo "$actual" | grep -qF "deepagents"; then break; fi
+      if echo "$actual" | grep -qF "openclaw"; then break; fi
     done
-    _check_image "$RT_LG_ID" "deepagents" "Runtime change langgraph→deepagents on restart"
+    _check_image "$RT_CX_ID" "openclaw" "Runtime change codex to openclaw on restart"
   else
     echo "  SKIP: Docker not available"
     SKIP=$((SKIP + 1))
@@ -228,7 +228,7 @@ else
 fi
 
 # Clean up runtime test workspaces
-for rt_id in $RT_CC_ID $RT_LG_ID $RT_CR_ID; do
+for rt_id in $RT_CC_ID $RT_CX_ID $RT_HM_ID; do
   curl -s -X DELETE "$BASE/workspaces/$rt_id?confirm=true" > /dev/null 2>&1
   sleep 0.3
 done

tests/e2e/test_model_slug.sh

@@ -2,12 +2,10 @@
 # Regression test for tests/e2e/lib/model_slug.sh.
 #
 # PR #2571 fixed a synth-E2E masking bug where MODEL_SLUG was hardcoded
-# to "openai/gpt-4o" (slash-form) but langgraph's init_chat_model needs
-# "openai:gpt-4o" (colon-form). Fix shipped as a per-runtime case
-# statement. Without this regression test, dropping any branch of the
-# case (or flipping a slug format) would silently revert behavior — the
-# E2E only fails as "Could not resolve authentication method" at the
-# very first message, after a successful tenant + workspace provision.
+# to "openai/gpt-4o" (slash-form). Without this regression test, dropping
+# any branch of the case (or flipping a slug format) would silently revert
+# behavior — the E2E only fails as "Could not resolve authentication method"
+# at the very first message, after a successful tenant + workspace provision.
 #
 # Each branch must FAIL the test if the dispatch behavior changes, not
 # just produce some non-empty string.
@@ -47,7 +45,7 @@ echo
 
 # ── Per-runtime branches (the load-bearing ones for synth-E2E) ──
 run_test "hermes → slash-form (derive-provider.sh contract)"       hermes      "openai/gpt-4o"
-run_test "langgraph → colon-form (init_chat_model contract)"       langgraph   "openai:gpt-4o"
+run_test "codex → slash-form fallback"                             codex       "openai/gpt-4o"
 run_test "claude-code → OAuth/default alias"                      claude-code "sonnet"
 
 got=$(unset E2E_MODEL_SLUG E2E_ANTHROPIC_API_KEY; E2E_MINIMAX_API_KEY="mx-test" pick_model_slug claude-code)
@@ -74,8 +72,8 @@ echo
 echo "Test: pick_model_slug — E2E_MODEL_SLUG override"
 echo
 
-got=$(E2E_MODEL_SLUG="anthropic:claude-opus-4-7" pick_model_slug langgraph)
-assert_eq "override beats langgraph default"                      "$got" "anthropic:claude-opus-4-7"
+got=$(E2E_MODEL_SLUG="anthropic:claude-opus-4-7" pick_model_slug codex)
+assert_eq "override beats codex default"                          "$got" "anthropic:claude-opus-4-7"
 
 got=$(E2E_MODEL_SLUG="custom/whatever" pick_model_slug hermes)
 assert_eq "override beats hermes default"                         "$got" "custom/whatever"
@@ -88,8 +86,8 @@ assert_eq "override beats claude-code default"                    "$got" "some-b
 # it because changing this behavior (e.g. via -v test) would silently
 # break the dispatch when an operator passes "" to clear an inherited
 # env var.
-got=$(E2E_MODEL_SLUG="" pick_model_slug langgraph)
-assert_eq "empty-string override falls through to dispatch"       "$got" "openai:gpt-4o"
+got=$(E2E_MODEL_SLUG="" pick_model_slug codex)
+assert_eq "empty-string override falls through to dispatch"       "$got" "openai/gpt-4o"
 
 echo
 echo "─────────────────────────────────────────────────"

tests/e2e/test_notify_attachments_e2e.sh

@@ -94,10 +94,10 @@ done
 
 # model is required at the Create boundary (CTO 2026-05-22 SSOT — see
 # feedback_workspace_model_required_no_platform_default_dynamic_credential_intake).
-# Body had no runtime → defaults to langgraph; pass the langgraph-compatible
-# default that the deleted DefaultModel("") would have returned.
+# Body has no runtime → defaults to claude-code; pass the matching model
+# that the workspace-creation contract now requires.
 R=$(curl -s -X POST "$BASE/workspaces" -H "Content-Type: application/json" \
-  -d '{"name":"Notify E2E","tier":1,"model":"anthropic:claude-opus-4-7"}')
+  -d '{"name":"Notify E2E","tier":1,"model":"sonnet"}')
 WSID=$(echo "$R" | python3 -c 'import json,sys;print(json.load(sys.stdin)["id"])' 2>/dev/null || true)
 [ -n "$WSID" ] || { echo "Failed to create workspace: $R"; exit 1; }
 echo "Created workspace $WSID"

tests/e2e/test_priority_runtimes_e2e.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# E2E test: every runtime template (8 total) works end-to-end.
+# E2E test: every maintained runtime template works end-to-end.
 #
 # Self-contained happy-path smoke per runtime. Provisions a fresh
 # workspace, waits for status=online, sends a real A2A message, and
@@ -7,13 +7,12 @@
 # extraction (and ongoing template work) can't silently break any
 # runtime.
 #
-# Runtimes covered: claude-code, hermes, langgraph, crewai, autogen,
-# deepagents, openclaw, gemini-cli. claude-code + hermes have unique
+# Runtimes covered: claude-code, codex, hermes, openclaw.
+# claude-code + hermes have unique
 # provisioning quirks (claude-code OAuth, hermes 15-min cold-boot)
 # and stay first-class with their own run_<runtime> functions; the
-# 5 OpenAI-backed runtimes share run_openai_runtime; gemini-cli has
-# its own block (Google AI key). Each phase skips cleanly if its
-# prerequisite secret is missing.
+# OpenAI-backed runtimes share run_openai_runtime. Each phase skips cleanly
+# if its prerequisite secret is missing.
 #
 # What this proves:
 #   1. Provisioning + container boot works for each runtime.
@@ -296,9 +295,8 @@ print(json.dumps({
 ####################################################################
 # Secondary runtimes — same provision/online/A2A loop, parametrized.
 ####################################################################
-# These 5 templates (langgraph, crewai, autogen, deepagents, openclaw)
-# all use OpenAI as their LLM provider in the default config and don't
-# need the hermes-specific HERMES_* secret block. Skip if no key.
+# Codex and OpenClaw use OpenAI as their LLM provider in this smoke and
+# don't need the hermes-specific HERMES_* secret block. Skip if no key.
 # claude-code + hermes stay first-class above because each has unique
 # provisioning quirks (claude-code OAuth, hermes cold-boot tolerance);
 # refactoring them into this generic loop would lose those guards.
@@ -361,67 +359,17 @@ print(json.dumps({
   fi
 }
 
-run_langgraph()  { run_openai_runtime "langgraph"  "langgraph"; }
-run_crewai()     { run_openai_runtime "crewai"     "crewai"; }
-run_autogen()    { run_openai_runtime "autogen"    "autogen"; }
-run_deepagents() { run_openai_runtime "deepagents" "deepagents"; }
+run_codex()      { run_openai_runtime "codex"      "codex"; }
 run_openclaw()   { run_openai_runtime "openclaw"   "openclaw"; }
 
-# gemini-cli wants a Google API key, not OpenAI. Skip if absent.
-run_gemini_cli() {
-  echo ""
-  echo "=== gemini-cli happy path ==="
-  if [ -z "${E2E_GEMINI_API_KEY:-}" ]; then
-    skip "E2E_GEMINI_API_KEY not set (gemini-cli needs Google AI key)"
-    return 0
-  fi
-  local secrets
-  secrets=$(python3 -c "
-import json, os
-print(json.dumps({'GEMINI_API_KEY': os.environ['E2E_GEMINI_API_KEY']}))
-")
-  local resp wsid
-  # model required (CTO 2026-05-22 SSOT) — gemini-cli routes via the gemini provider.
-  resp=$(curl -s -X POST "$BASE/workspaces" -H "Content-Type: application/json" \
-    -d "{\"name\":\"Priority E2E (gemini-cli)\",\"runtime\":\"gemini-cli\",\"model\":\"gemini-2.0-flash\",\"tier\":1,\"secrets\":$secrets}")
-  wsid=$(echo "$resp" | python3 -c 'import json,sys;print(json.load(sys.stdin).get("id",""))') || true
-  if [ -z "$wsid" ]; then fail "create gemini-cli workspace" "$resp"; return 0; fi
-  CREATED_WSIDS+=("$wsid")
-  echo "  workspace=$wsid"
-  local final
-  final=$(wait_for_status "$wsid" "online failed" 240) || true
-  if [ "$final" != "online" ]; then
-    fail "gemini-cli workspace reaches online" "final status: $final"
-    return 0
-  fi
-  pass "gemini-cli workspace reaches online"
-  local token; token=$(e2e_mint_test_token "$wsid")
-  if [ -z "$token" ]; then fail "mint gemini-cli test token" "no token"; return 0; fi
-  local reply
-  if reply=$(send_test_prompt "$wsid" "$token"); then
-    if echo "$reply" | grep -q "PONG"; then
-      pass "gemini-cli reply contains PONG"
-    else
-      pass "gemini-cli reply non-empty (first 80 chars: ${reply:0:80})"
-    fi
-    assert_activity_logged "gemini-cli" "$wsid" "$token"
-  else
-    fail "gemini-cli reply" "${reply:-<empty or error>}"
-  fi
-}
-
-WANT="${E2E_RUNTIMES:-claude-code hermes}"
+WANT="${E2E_RUNTIMES:-claude-code codex hermes openclaw}"
 for r in $WANT; do
   case "$r" in
     claude-code) run_claude_code ;;
+    codex)       run_codex ;;
     hermes)      run_hermes ;;
-    langgraph)   run_langgraph ;;
-    crewai)      run_crewai ;;
-    autogen)     run_autogen ;;
-    deepagents)  run_deepagents ;;
     openclaw)    run_openclaw ;;
-    gemini-cli)  run_gemini_cli ;;
-    all)         run_claude_code; run_hermes; run_langgraph; run_crewai; run_autogen; run_deepagents; run_openclaw; run_gemini_cli ;;
+    all)         run_claude_code; run_codex; run_hermes; run_openclaw ;;
     *) echo "unknown runtime in E2E_RUNTIMES: $r" >&2; exit 2 ;;
   esac
 done

tests/harness/seed.sh

@@ -27,9 +27,9 @@ create_workspace() {
     local tenant="$1" name="$2" tier="$3" parent="${4:-}"
     local body
     if [ -n "$parent" ]; then
-        body="{\"name\":\"$name\",\"tier\":$tier,\"parent_id\":\"$parent\",\"runtime\":\"langgraph\"}"
+        body="{\"name\":\"$name\",\"tier\":$tier,\"parent_id\":\"$parent\",\"runtime\":\"claude-code\",\"model\":\"sonnet\"}"
     else
-        body="{\"name\":\"$name\",\"tier\":$tier,\"runtime\":\"langgraph\"}"
+        body="{\"name\":\"$name\",\"tier\":$tier,\"runtime\":\"claude-code\",\"model\":\"sonnet\"}"
     fi
     local id
     if [ "$tenant" = "alpha" ]; then

workspace-server/internal/bundle/importer.go

@@ -60,8 +60,8 @@ func Import(
 	// Build config files in memory for the provisioner
 	configFiles := buildBundleConfigFiles(b)
 
-	// Extract runtime from config.yaml in the bundle
-	bundleRuntime := "langgraph"
+	// Extract runtime from config.yaml in the bundle.
+	bundleRuntime := "claude-code"
 	if configYaml, ok := b.Prompts["config.yaml"]; ok {
 		for _, line := range strings.Split(configYaml, "\n") {
 			line = strings.TrimSpace(line)

workspace-server/internal/handlers/a2a_proxy.go

@@ -705,7 +705,7 @@ func (h *WorkspaceHandler) resolveAgentURL(ctx context.Context, workspaceID stri
 	if strings.HasPrefix(agentURL, "http://127.0.0.1:") && h.provisioner != nil && platformInDocker {
 		var wsRuntime string
 		if err := db.DB.QueryRowContext(ctx,
-			`SELECT COALESCE(runtime, 'langgraph') FROM workspaces WHERE id = $1`,
+			`SELECT COALESCE(runtime, 'claude-code') FROM workspaces WHERE id = $1`,
 			workspaceID,
 		).Scan(&wsRuntime); err != nil {
 			log.Printf("ProxyA2A: runtime lookup before Docker URL rewrite failed for %s: %v", workspaceID, err)

workspace-server/internal/handlers/a2a_proxy_helpers.go

@@ -169,7 +169,7 @@ func (h *WorkspaceHandler) handleA2ADispatchError(ctx context.Context, workspace
 // canvas-chat-to-dead-workspace incident traces to exactly this gap.
 func (h *WorkspaceHandler) maybeMarkContainerDead(ctx context.Context, workspaceID string) bool {
 	var wsRuntime string
-	db.DB.QueryRowContext(ctx, `SELECT COALESCE(runtime, 'langgraph') FROM workspaces WHERE id = $1`, workspaceID).Scan(&wsRuntime)
+	db.DB.QueryRowContext(ctx, `SELECT COALESCE(runtime, 'claude-code') FROM workspaces WHERE id = $1`, workspaceID).Scan(&wsRuntime)
 	if isExternalLikeRuntime(wsRuntime) {
 		return false
 	}

workspace-server/internal/handlers/a2a_proxy_test.go

@@ -280,7 +280,7 @@ func TestProxyA2A_Upstream502_TriggersContainerDeadCheck(t *testing.T) {
 	mock.ExpectExec("INSERT INTO activity_logs").
 		WillReturnResult(sqlmock.NewResult(0, 1))
 	// maybeMarkContainerDead's runtime lookup, then the offline-flip UPDATE.
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-tunnel-dead").
 		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("hermes"))
 	mock.ExpectExec(`UPDATE workspaces SET status =`).
@@ -340,7 +340,7 @@ func TestProxyA2A_Upstream502_AliveAgent_PropagatesAsIs(t *testing.T) {
 	expectBudgetCheck(mock, "ws-alive-502")
 	mock.ExpectExec("INSERT INTO activity_logs").WillReturnResult(sqlmock.NewResult(0, 1))
 	// IsRunning runtime lookup runs but no UPDATE follows (running=true).
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-alive-502").
 		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("hermes"))
 
@@ -1954,9 +1954,9 @@ func TestMaybeMarkContainerDead_NilProvisioner(t *testing.T) {
 	setupTestRedis(t)
 	handler := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())
 
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-nilprov").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("langgraph"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("claude-code"))
 
 	if got := handler.maybeMarkContainerDead(context.Background(), "ws-nilprov"); got {
 		t.Error("expected false when provisioner is nil")
@@ -1977,7 +1977,7 @@ func TestMaybeMarkContainerDead_CPOnly_NotRunning(t *testing.T) {
 	cp := &fakeCPProv{running: false}
 	handler.SetCPProvisioner(cp)
 
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-saas-dead").
 		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("hermes"))
 	mock.ExpectExec(`UPDATE workspaces SET status =`).
@@ -2006,7 +2006,7 @@ func TestMaybeMarkContainerDead_CPOnly_Running(t *testing.T) {
 	cp := &fakeCPProv{running: true}
 	handler.SetCPProvisioner(cp)
 
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-saas-alive").
 		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("hermes"))
 
@@ -2107,7 +2107,7 @@ func TestMaybeMarkContainerDead_ExternalRuntime(t *testing.T) {
 	setupTestRedis(t)
 	handler := NewWorkspaceHandler(newTestBroadcaster(), nil, "http://localhost:8080", t.TempDir())
 
-	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime, 'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-ext").
 		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("external"))
 

workspace-server/internal/handlers/discovery.go

@@ -128,7 +128,7 @@ func discoverHostPeer(ctx context.Context, c *gin.Context, targetID string) {
 // of `callerID` and writes the JSON response (or an appropriate 404/503 error).
 func discoverWorkspacePeer(ctx context.Context, c *gin.Context, callerID, targetID string) {
 	var wsName, wsRuntime string
-	db.DB.QueryRowContext(ctx, `SELECT COALESCE(name,''), COALESCE(runtime,'langgraph') FROM workspaces WHERE id = $1`, targetID).Scan(&wsName, &wsRuntime)
+	db.DB.QueryRowContext(ctx, `SELECT COALESCE(name,''), COALESCE(runtime,'claude-code') FROM workspaces WHERE id = $1`, targetID).Scan(&wsName, &wsRuntime)
 
 	// External workspaces: return their registered URL.
 	// Rewrite 127.0.0.1/localhost → host.docker.internal ONLY when the
@@ -180,7 +180,7 @@ func writeExternalWorkspaceURL(ctx context.Context, c *gin.Context, callerID, ta
 	}
 	outURL := wsURL
 	var callerRuntime string
-	db.DB.QueryRowContext(ctx, `SELECT COALESCE(runtime,'langgraph') FROM workspaces WHERE id = $1`, callerID).Scan(&callerRuntime)
+	db.DB.QueryRowContext(ctx, `SELECT COALESCE(runtime,'claude-code') FROM workspaces WHERE id = $1`, callerID).Scan(&callerRuntime)
 	if !isExternalLikeRuntime(callerRuntime) {
 		outURL = strings.Replace(outURL, "127.0.0.1", "host.docker.internal", 1)
 		outURL = strings.Replace(outURL, "localhost", "host.docker.internal", 1)
@@ -203,8 +203,8 @@ func writeExternalWorkspaceURL(ctx context.Context, c *gin.Context, callerID, ta
 
 // Peers handles GET /registry/:id/peers
 //
-// Optional ``?q=<substring>`` filters the result by case-insensitive
-// substring match against ``name`` or ``role`` (#1038). Filtering is done
+// Optional “?q=<substring>“ filters the result by case-insensitive
+// substring match against “name“ or “role“ (#1038). Filtering is done
 // in Go after the DB read — keeps the SQL identical to the no-filter path
 // (no injection risk, no DB-driver collation surprises) at the cost of
 // loading the unfiltered set first. Acceptable because the peer set is
@@ -301,8 +301,8 @@ func (h *DiscoveryHandler) Peers(c *gin.Context) {
 	c.JSON(http.StatusOK, peers)
 }
 
-// excludeSelfFromPeers strips any peer entry whose ``id`` equals
-// ``workspaceID`` (the caller's own row). Final-line defense for #383
+// excludeSelfFromPeers strips any peer entry whose “id“ equals
+// “workspaceID“ (the caller's own row). Final-line defense for #383
 // (self-delegation 400-loop on external workspaces): a peer-list that
 // includes the requester's own row is the root mechanism by which an
 // agent ends up delegating to itself. The pre-DB filters in Peers
@@ -337,8 +337,8 @@ func filterPeersByQuery(peers []map[string]interface{}, q string) []map[string]i
 	needle := strings.ToLower(q)
 	out := make([]map[string]interface{}, 0, len(peers))
 	for _, p := range peers {
-		name, _ := p["name"].(string)  // nil → "" — safe on empty-role rows
-		role, _ := p["role"].(string)  // nil → "" — queryPeerMaps sets nil when DB role is empty
+		name, _ := p["name"].(string) // nil → "" — safe on empty-role rows
+		role, _ := p["role"].(string) // nil → "" — queryPeerMaps sets nil when DB role is empty
 		if strings.Contains(strings.ToLower(name), needle) ||
 			strings.Contains(strings.ToLower(role), needle) {
 			out = append(out, p)

workspace-server/internal/handlers/discovery_test.go

@@ -404,11 +404,11 @@ func TestPeers_Q_NoMatches_RawBodyIsArrayNotNull(t *testing.T) {
 // role and asserts no panic + correct filter behaviour.
 func TestFilterPeersByQuery_NilRoleRegression(t *testing.T) {
 	cases := []struct {
-		name       string
-		peers      []map[string]interface{}
-		q          string
-		wantLen    int
-		wantIDs    []string
+		name    string
+		peers   []map[string]interface{}
+		q       string
+		wantLen int
+		wantIDs []string
 	}{
 		{
 			name: "nil role matches on name",
@@ -555,7 +555,7 @@ func TestDiscoverWorkspacePeer_Online(t *testing.T) {
 	setupTestRedis(t)
 
 	// name/runtime lookup → non-external
-	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-online").
 		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("Target", "langgraph"))
 	// No cached internal URL → DB status lookup → online
@@ -583,7 +583,7 @@ func TestDiscoverWorkspacePeer_NotFound(t *testing.T) {
 	mock := setupTestDB(t)
 	setupTestRedis(t)
 
-	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-missing").
 		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("", "langgraph"))
 	mock.ExpectQuery(`SELECT status FROM workspaces WHERE id =`).
@@ -605,14 +605,14 @@ func TestDiscoverWorkspacePeer_ExternalRuntime_HandledByExternalURL(t *testing.T
 	mock := setupTestDB(t)
 	setupTestRedis(t)
 
-	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-ext").
 		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("Ext", "external"))
 	// writeExternalWorkspaceURL's two queries
 	mock.ExpectQuery(`SELECT COALESCE\(url,''\) FROM workspaces WHERE id =`).
 		WithArgs("ws-ext").
 		WillReturnRows(sqlmock.NewRows([]string{"url"}).AddRow("http://external.example"))
-	mock.ExpectQuery(`SELECT COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-caller").
 		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("external"))
 
@@ -630,7 +630,7 @@ func TestDiscoverWorkspacePeer_CachedInternalURLHit(t *testing.T) {
 	mock := setupTestDB(t)
 	mr := setupTestRedis(t)
 
-	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-cached").
 		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("Cached", "langgraph"))
 	mr.Set("ws:ws-cached:internal_url", "http://ws-cached:8000")
@@ -654,7 +654,7 @@ func TestDiscoverWorkspacePeer_NotReachable(t *testing.T) {
 	mock := setupTestDB(t)
 	setupTestRedis(t)
 
-	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(name,''\), COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-paused").
 		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("Paused", "langgraph"))
 	mock.ExpectQuery(`SELECT status FROM workspaces WHERE id =`).
@@ -681,9 +681,9 @@ func TestWriteExternalWorkspaceURL_Success(t *testing.T) {
 	mock.ExpectQuery(`SELECT COALESCE\(url,''\) FROM workspaces WHERE id =`).
 		WithArgs("ws-ext").
 		WillReturnRows(sqlmock.NewRows([]string{"url"}).AddRow("http://external.example/a2a"))
-	mock.ExpectQuery(`SELECT COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-caller").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("langgraph"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("claude-code"))
 
 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -731,9 +731,9 @@ func TestWriteExternalWorkspaceURL_RewritesLocalhostForDockerCaller(t *testing.T
 		WithArgs("ws-ext").
 		WillReturnRows(sqlmock.NewRows([]string{"url"}).AddRow("http://127.0.0.1:8000/a2a"))
 	// non-external caller runtime → rewrite enabled
-	mock.ExpectQuery(`SELECT COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-caller").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("langgraph"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("claude-code"))
 
 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
@@ -818,9 +818,9 @@ func TestWriteExternalWorkspaceURL_RejectsMetadataIPURL(t *testing.T) {
 		WillReturnRows(sqlmock.NewRows([]string{"url"}).
 			AddRow("http://169.254.169.254/computeMetadata/v1/"))
 	// callerRuntime lookup happens before isSafeURL — must mock it.
-	mock.ExpectQuery(`SELECT COALESCE\(runtime,'langgraph'\) FROM workspaces WHERE id =`).
+	mock.ExpectQuery(`SELECT COALESCE\(runtime,'claude-code'\) FROM workspaces WHERE id =`).
 		WithArgs("ws-caller").
-		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("langgraph"))
+		WillReturnRows(sqlmock.NewRows([]string{"runtime"}).AddRow("claude-code"))
 
 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)

workspace-server/internal/handlers/handlers_additional_test.go

@@ -33,9 +33,11 @@ func TestWorkspaceCreate_WithParentID(t *testing.T) {
 	// Default tier is 3 (Privileged) — see workspace.go create-handler comment.
 	// delivery_mode defaults to "push" when payload omits it (#2339).
 	mock.ExpectExec("INSERT INTO workspaces").
-		WithArgs(sqlmock.AnyArg(), "Child Agent", nil, 3, "langgraph", &parentID, nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
+		WithArgs(sqlmock.AnyArg(), "Child Agent", nil, 3, "claude-code", &parentID, nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
 		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectCommit()
+	mock.ExpectExec("INSERT INTO workspace_secrets").
+		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectExec("INSERT INTO canvas_layouts").
 		WithArgs(sqlmock.AnyArg(), float64(0), float64(0)).
 		WillReturnResult(sqlmock.NewResult(0, 1))
@@ -104,7 +106,7 @@ func TestWorkspaceCreate_MissingName(t *testing.T) {
 
 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)
-	body := `{"tier":1,"runtime":"langgraph"}`
+	body := `{"tier":1,"runtime":"claude-code"}`
 	c.Request = httptest.NewRequest("POST", "/workspaces", bytes.NewBufferString(body))
 	c.Request.Header.Set("Content-Type", "application/json")
 
@@ -242,7 +244,7 @@ func TestWorkspaceList_WithData(t *testing.T) {
 	}
 	rows := sqlmock.NewRows(columns).
 		AddRow("ws-1", "Agent One", "worker", 1, "online", []byte(`{"name":"agent1"}`), "http://localhost:8001",
-			nil, 3, 1, 0.02, "", 7200, "processing", "langgraph", "", 10.0, 20.0, false, nil, int64(0), false, true, []byte(`{}`)).
+			nil, 3, 1, 0.02, "", 7200, "processing", "claude-code", "", 10.0, 20.0, false, nil, int64(0), false, true, []byte(`{}`)).
 		AddRow("ws-2", "Agent Two", "", 2, "degraded", []byte("null"), "",
 			nil, 0, 1, 0.6, "timeout", 100, "", "claude-code", "", 50.0, 60.0, true, nil, int64(0), false, true, []byte(`{}`))
 
@@ -585,7 +587,7 @@ func TestDiscover_TargetOffline(t *testing.T) {
 	// Name + runtime lookup (discovery now queries both)
 	mock.ExpectQuery("SELECT COALESCE").
 		WithArgs("ws-off").
-		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("Offline Agent", "langgraph"))
+		WillReturnRows(sqlmock.NewRows([]string{"name", "runtime"}).AddRow("Offline Agent", "claude-code"))
 
 	// No cached internal URL → falls to DB status check → offline
 	mock.ExpectQuery("SELECT status FROM workspaces WHERE id =").
@@ -917,7 +919,7 @@ func TestRestart_ParentPaused(t *testing.T) {
 	mock.ExpectQuery("SELECT status, name, tier").
 		WithArgs("dddddddd-0001-0000-0000-000000000000").
 		WillReturnRows(sqlmock.NewRows([]string{"status", "name", "tier", "runtime"}).
-			AddRow("offline", "Child Agent", 1, "langgraph"))
+			AddRow("offline", "Child Agent", 1, "claude-code"))
 
 	// isParentPaused: get parent_id
 	mock.ExpectQuery("SELECT parent_id FROM workspaces WHERE id").
@@ -960,7 +962,7 @@ func TestRestart_ProvisionerNil(t *testing.T) {
 	mock.ExpectQuery("SELECT status, name, tier").
 		WithArgs("ws-noprov").
 		WillReturnRows(sqlmock.NewRows([]string{"status", "name", "tier", "runtime"}).
-			AddRow("offline", "Agent", 1, "langgraph"))
+			AddRow("offline", "Agent", 1, "claude-code"))
 
 	// isParentPaused: no parent
 	mock.ExpectQuery("SELECT parent_id FROM workspaces WHERE id").
@@ -1195,7 +1197,7 @@ func TestResume_ProvisionerNil(t *testing.T) {
 	mock.ExpectQuery("SELECT name, tier").
 		WithArgs("ws-resume-noprov").
 		WillReturnRows(sqlmock.NewRows([]string{"name", "tier", "runtime"}).
-			AddRow("Paused Agent", 1, "langgraph"))
+			AddRow("Paused Agent", 1, "claude-code"))
 
 	w := httptest.NewRecorder()
 	c, _ := gin.CreateTestContext(w)

workspace-server/internal/handlers/handlers_test.go

@@ -368,11 +368,13 @@ func TestWorkspaceCreate(t *testing.T) {
 	// Default tier is 3 (Privileged) — see workspace.go create-handler comment.
 	// delivery_mode defaults to "push" when payload omits it (#2339).
 	mock.ExpectExec("INSERT INTO workspaces").
-		WithArgs(sqlmock.AnyArg(), "Test Agent", nil, 3, "langgraph", (*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
+		WithArgs(sqlmock.AnyArg(), "Test Agent", nil, 3, "claude-code", (*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
 		WillReturnResult(sqlmock.NewResult(0, 1))
 
 	// Expect transaction commit (no secrets in this payload)
 	mock.ExpectCommit()
+	mock.ExpectExec("INSERT INTO workspace_secrets").
+		WillReturnResult(sqlmock.NewResult(0, 1))
 
 	// Expect canvas_layouts INSERT
 	mock.ExpectExec("INSERT INTO canvas_layouts").
@@ -389,8 +391,8 @@ func TestWorkspaceCreate(t *testing.T) {
 	// Note: model is now required at the Create boundary (CTO 2026-05-22
 	// SSOT directive — see feedback_workspace_model_required_no_platform_default_dynamic_credential_intake
 	// and TestCreate_ModelRequired_Returns422). This test happens to take
-	// the bare-defaults path (no template, no runtime → langgraph), so
-	// the body must declare an explicit model. Using a langgraph-compatible
+	// the bare-defaults path (no template, no runtime → claude-code), so
+	// the body must declare an explicit model. Using a claude-code-compatible
 	// id; the test doesn't exercise model semantics beyond presence.
 	body := `{"name":"Test Agent","model":"anthropic:claude-opus-4-7","canvas":{"x":100,"y":200}}`
 	c.Request = httptest.NewRequest("POST", "/workspaces", bytes.NewBufferString(body))
@@ -462,7 +464,7 @@ func TestWorkspaceList(t *testing.T) {
 		AddRow("ws-1", "Agent One", "worker", 1, "online", []byte("null"), "http://localhost:8001",
 			nil, 0, 1, 0.0, "", 100, "", "claude-code", "", 10.0, 20.0, false, nil, int64(0), false, true, []byte(`{}`)).
 		AddRow("ws-2", "Agent Two", "manager", 2, "provisioning", []byte("null"), "",
-			nil, 0, 1, 0.0, "", 0, "", "langgraph", "", 50.0, 60.0, false, nil, int64(0), false, true, []byte(`{}`))
+			nil, 0, 1, 0.0, "", 0, "", "claude-code", "", 50.0, 60.0, false, nil, int64(0), false, true, []byte(`{}`))
 
 	mock.ExpectQuery("SELECT w.id, w.name").
 		WillReturnRows(rows)
@@ -1182,7 +1184,7 @@ func TestWorkspaceGet_CurrentTask(t *testing.T) {
 		WithArgs("dddddddd-0004-0000-0000-000000000000").
 		WillReturnRows(sqlmock.NewRows(columns).AddRow(
 			"dddddddd-0004-0000-0000-000000000000", "Task Worker", "worker", 1, "online", []byte("null"), "http://localhost:9000",
-			nil, 2, 1, 0.0, "", 300, "Analyzing document", "langgraph", "", 10.0, 20.0, false,
+			nil, 2, 1, 0.0, "", 300, "Analyzing document", "claude-code", "", 10.0, 20.0, false,
 			nil, int64(0), false, true, []byte(`{}`),
 		))
 

workspace-server/internal/handlers/org_import.go

@@ -62,7 +62,7 @@ func (h *OrgHandler) createWorkspaceTree(ws OrgWorkspace, parentID *string, absX
 		runtime = defaults.Runtime
 	}
 	if runtime == "" {
-		runtime = "langgraph"
+		runtime = "claude-code"
 	}
 	model := ws.Model
 	if model == "" {

workspace-server/internal/handlers/plugins_install_eic_test.go

@@ -119,7 +119,6 @@ func TestHostPluginPath_PerRuntime(t *testing.T) {
 	}{
 		{"claude-code", "browser-automation", "/configs/plugins/browser-automation"},
 		{"hermes", "browser-automation", "/home/ubuntu/.hermes/plugins/browser-automation"},
-		{"langgraph", "browser-automation", "/opt/configs/plugins/browser-automation"},
 		// Unknown / empty runtime falls back to /configs (containerized
 		// user-data layout) so a future runtime added to workspaces table
 		// without a workspaceFilePathPrefix entry doesn't blow up the

workspace-server/internal/handlers/restart_template.go

@@ -89,9 +89,9 @@ func resolveRestartTemplate(configsDir, wsName, dbRuntime string, body restartTe
 	// host, load an arbitrary host directory as the workspace template.
 	//
 	// sanitizeRuntime applies an allowlist of known runtimes; any unknown
-	// value (including traversal strings) is remapped to "langgraph".  The
+	// value (including traversal strings) is remapped to "claude-code".  The
 	// attacker cannot choose an arbitrary host path — they can at most
-	// trigger application of the langgraph-default template.
+	// trigger application of the claude-code-default template.
 	if body.ApplyTemplate && dbRuntime != "" {
 		safeRuntime := sanitizeRuntime(dbRuntime)
 		runtimeTemplate := filepath.Join(configsDir, safeRuntime+"-default")

workspace-server/internal/handlers/template_files_eic.go

@@ -48,13 +48,10 @@ import (
 var workspaceFilePathPrefix = map[string]string{
 	"hermes":      "/home/ubuntu/.hermes",
 	"claude-code": "/configs",
-	"langgraph":   "/opt/configs",
 	"external":    "/opt/configs",
 	// Default for unknown / future runtimes is /configs — matches the
-	// containerized user-data layout. The `langgraph` / `external`
-	// entries pre-date the unified user-data path and are retained
-	// until a migration audit confirms what the running tenants of
-	// those runtimes actually have on disk.
+	// containerized user-data layout. The `external` entry pre-dates the
+	// unified user-data path and does not map to a spawned runtime.
 }
 
 // resolveWorkspaceFilePath translates (runtime, root, relPath) into an

workspace-server/internal/handlers/template_files_eic_test.go

@@ -32,7 +32,6 @@ func TestResolveWorkspaceFilePath_RuntimeIndirection(t *testing.T) {
 		// and would 500 with EACCES on save (the bug that motivated this gate).
 		{"claude-code", "/configs", "config.yaml", "/configs/config.yaml"},
 		{"CLAUDE-CODE", "/configs", "config.yaml", "/configs/config.yaml"}, // case-insensitive
-		{"langgraph", "/configs", "config.yaml", "/opt/configs/config.yaml"},
 		{"external", "/configs", "skills.json", "/opt/configs/skills.json"},
 		{"", "/configs", "config.yaml", "/configs/config.yaml"},        // empty runtime → default
 		{"unknown", "/configs", "config.yaml", "/configs/config.yaml"}, // unknown → default
@@ -70,7 +69,7 @@ func TestResolveWorkspaceFilePath_LiteralRoots(t *testing.T) {
 		// universal Linux path, not a managed-config indirection.
 		{"hermes", "/home", "ubuntu/.bashrc", "/home/ubuntu/.bashrc"},
 		{"claude-code", "/home", "ubuntu/notes.md", "/home/ubuntu/notes.md"},
-		{"langgraph", "/home", "ubuntu/x", "/home/ubuntu/x"},
+		{"codex", "/home", "ubuntu/x", "/home/ubuntu/x"},
 		// /workspace and /plugins are also literal — runtime is ignored.
 		{"hermes", "/workspace", "src/main.go", "/workspace/src/main.go"},
 		{"claude-code", "/plugins", "p/manifest.yaml", "/plugins/p/manifest.yaml"},

workspace-server/internal/handlers/workspace.go

@@ -273,12 +273,12 @@ func (h *WorkspaceHandler) Create(c *gin.Context) {
 	// runtimeExplicitlyRequested is true when the caller expressed intent for
 	// a SPECIFIC runtime — either by passing `runtime` directly, or by naming
 	// a `template` (a template encodes a runtime). When true, we must NOT
-	// silently fall back to langgraph if that intent can't be honored: that
+	// silently fall back to the default runtime if that intent can't be honored: that
 	// is the molecule-controlplane#188 / #184 contract violation (caller asks
-	// for codex/claude-code, gets a langgraph workspace, 201, no error — a
+	// for codex/hermes/openclaw, gets a default-runtime workspace, 201, no error — a
 	// false success). #188 mandates fail-closed (error+notify) on mismatch,
 	// not an advisory degrade. The legitimate "no template, no runtime →
-	// langgraph default" path (bare {"name":...}) is unaffected.
+	// default-runtime path (bare {"name":...}) is unaffected.
 	runtimeExplicitlyRequested := payload.Runtime != "" || payload.Template != ""
 	templateRuntimeResolved := payload.Runtime != ""
 	if payload.Template != "" && (payload.Runtime == "" || payload.Model == "") {
@@ -332,15 +332,15 @@ func (h *WorkspaceHandler) Create(c *gin.Context) {
 	// intent for a specific runtime (passed `runtime`, or named a `template`)
 	// but we could NOT resolve a concrete runtime from it (template's
 	// config.yaml unreadable, or it has no `runtime:` key), DO NOT silently
-	// substitute langgraph and return 201 — that is the silent contract
+	// substitute the default runtime and return 201 — that is the silent contract
 	// violation that produced 5/5 wrong workspaces and a false codex E2E pass.
 	// Return 422 so the caller learns the requested runtime was not honored.
 	// The platform-side CP fix (controlplane#188) is the sibling gate; this
 	// closes the ws-server `Create` boundary the product UI actually hits.
 	if payload.Runtime == "" && runtimeExplicitlyRequested && !templateRuntimeResolved {
-		log.Printf("Create: FAIL-CLOSED (controlplane#188) — template=%q requested but runtime could not be resolved; refusing silent langgraph fallback", payload.Template)
+		log.Printf("Create: FAIL-CLOSED (controlplane#188) — template=%q requested but runtime could not be resolved; refusing silent default-runtime fallback", payload.Template)
 		c.JSON(http.StatusUnprocessableEntity, gin.H{
-			"error":    "runtime could not be resolved from the requested template; refusing to silently provision langgraph (controlplane#188). Pass an explicit \"runtime\", or use a template whose config.yaml declares one.",
+			"error":    "runtime could not be resolved from the requested template; refusing to silently provision the default runtime (controlplane#188). Pass an explicit \"runtime\", or use a template whose config.yaml declares one.",
 			"template": payload.Template,
 			"code":     "RUNTIME_UNRESOLVED",
 		})
@@ -348,8 +348,8 @@ func (h *WorkspaceHandler) Create(c *gin.Context) {
 	}
 	if payload.Runtime == "" {
 		// Legitimate default path: no template AND no runtime requested
-		// (bare {"name":...}) — langgraph is the intended default here.
-		payload.Runtime = "langgraph"
+		// (bare {"name":...}) — claude-code is the intended default here.
+		payload.Runtime = "claude-code"
 	}
 
 	// SSOT (CTO 2026-05-22, feedback_workspace_model_required_no_platform_default_dynamic_credential_intake):
@@ -851,7 +851,7 @@ const workspaceListQuery = `
 		   w.parent_id, w.active_tasks, COALESCE(w.max_concurrent_tasks, 1),
 		   w.last_error_rate,
 		   COALESCE(w.last_sample_error, ''), w.uptime_seconds,
-		   COALESCE(w.current_task, ''), COALESCE(w.runtime, 'langgraph'),
+		   COALESCE(w.current_task, ''), COALESCE(w.runtime, 'claude-code'),
 		   COALESCE(w.workspace_dir, ''),
 		   COALESCE(cl.x, 0), COALESCE(cl.y, 0), COALESCE(cl.collapsed, false),
 		   w.budget_limit, COALESCE(w.monthly_spend, 0),
@@ -913,7 +913,7 @@ func (h *WorkspaceHandler) Get(c *gin.Context) {
 			   w.parent_id, w.active_tasks, COALESCE(w.max_concurrent_tasks, 1),
 			   w.last_error_rate,
 			   COALESCE(w.last_sample_error, ''), w.uptime_seconds,
-			   COALESCE(w.current_task, ''), COALESCE(w.runtime, 'langgraph'),
+			   COALESCE(w.current_task, ''), COALESCE(w.runtime, 'claude-code'),
 			   COALESCE(w.workspace_dir, ''),
 			   COALESCE(cl.x, 0), COALESCE(cl.y, 0), COALESCE(cl.collapsed, false),
 			   w.budget_limit, COALESCE(w.monthly_spend, 0),

workspace-server/internal/handlers/workspace_budget_test.go

@@ -51,7 +51,7 @@ func TestWorkspaceBudget_Get_NilLimit(t *testing.T) {
 		WillReturnRows(sqlmock.NewRows(wsColumns).
 			AddRow("dddddddd-0005-0000-0000-000000000000", "Free Agent", "worker", 1, "online",
 				[]byte(`{}`), "http://localhost:9001",
-				nil, 0, 1, 0.0, "", 0, "", "langgraph", "",
+				nil, 0, 1, 0.0, "", 0, "", "claude-code", "",
 				0.0, 0.0, false,
 				nil,   // budget_limit NULL
 				0,     // monthly_spend 0
@@ -97,7 +97,7 @@ func TestWorkspaceBudget_Get_WithLimit(t *testing.T) {
 		WillReturnRows(sqlmock.NewRows(wsColumns).
 			AddRow("dddddddd-0006-0000-0000-000000000000", "Capped Agent", "worker", 1, "online",
 				[]byte(`{}`), "http://localhost:9002",
-				nil, 0, 1, 0.0, "", 0, "", "langgraph", "",
+				nil, 0, 1, 0.0, "", 0, "", "claude-code", "",
 				0.0, 0.0, false,
 				int64(500),  // budget_limit = $5.00 in DB
 				int64(123),  // monthly_spend = $1.23 in DB
@@ -151,7 +151,7 @@ func TestWorkspaceBudget_Create_WithLimit(t *testing.T) {
 			"Budgeted Agent",                 // name
 			nil,                              // role
 			3,                                // tier (default, workspace.go create-handler)
-			"langgraph",                      // runtime
+			"claude-code",                    // runtime
 			(*string)(nil),                   // parent_id
 			nil,                              // workspace_dir
 			"none",                           // workspace_access
@@ -161,6 +161,8 @@ func TestWorkspaceBudget_Create_WithLimit(t *testing.T) {
 		).
 		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectCommit()
+	mock.ExpectExec("INSERT INTO workspace_secrets").
+		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectExec("INSERT INTO canvas_layouts").
 		WithArgs(sqlmock.AnyArg(), float64(0), float64(0)).
 		WillReturnResult(sqlmock.NewResult(0, 1))

workspace-server/internal/handlers/workspace_provision.go

@@ -515,7 +515,7 @@ func configDirName(workspaceID string) string {
 }
 
 // knownRuntimes is the allowlist of runtime strings the provisioner will
-// accept. Unknown values are coerced to the default ("langgraph") instead
+// accept. Unknown values are coerced to the default ("claude-code") instead
 // of being splatted into filepath.Join + config.yaml templating, which
 // closes both the YAML-injection vector (#241) where an attacker could
 // smuggle `initial_prompt: run id && curl …` through a crafted runtime

workspace-server/internal/handlers/workspace_restart.go

@@ -212,7 +212,7 @@ func (h *WorkspaceHandler) Restart(c *gin.Context) {
 	var status, wsName, dbRuntime string
 	var tier int
 	err := db.DB.QueryRowContext(ctx,
-		`SELECT status, name, tier, COALESCE(runtime, 'langgraph') FROM workspaces WHERE id = $1`, id,
+		`SELECT status, name, tier, COALESCE(runtime, 'claude-code') FROM workspaces WHERE id = $1`, id,
 	).Scan(&status, &wsName, &tier, &dbRuntime)
 	if err == sql.ErrNoRows {
 		c.JSON(http.StatusNotFound, gin.H{"error": "workspace not found"})
@@ -744,7 +744,7 @@ func (h *WorkspaceHandler) runRestartCycle(workspaceID string) {
 	var wsName, status, dbRuntime string
 	var tier int
 	err := db.DB.QueryRowContext(ctx,
-		`SELECT name, status, tier, COALESCE(runtime, 'langgraph') FROM workspaces WHERE id = $1 AND status NOT IN ('removed', 'paused', 'hibernated')`, workspaceID,
+		`SELECT name, status, tier, COALESCE(runtime, 'claude-code') FROM workspaces WHERE id = $1 AND status NOT IN ('removed', 'paused', 'hibernated')`, workspaceID,
 	).Scan(&wsName, &status, &tier, &dbRuntime)
 	if err != nil {
 		return // includes paused/hibernated — don't auto-restart those
@@ -897,7 +897,7 @@ func (h *WorkspaceHandler) Resume(c *gin.Context) {
 	var wsName, dbRuntime string
 	var tier int
 	err := db.DB.QueryRowContext(ctx,
-		`SELECT name, tier, COALESCE(runtime, 'langgraph') FROM workspaces WHERE id = $1 AND status = 'paused'`, id,
+		`SELECT name, tier, COALESCE(runtime, 'claude-code') FROM workspaces WHERE id = $1 AND status = 'paused'`, id,
 	).Scan(&wsName, &tier, &dbRuntime)
 	if err == sql.ErrNoRows {
 		c.JSON(http.StatusNotFound, gin.H{"error": "workspace not found or not paused"})
@@ -930,9 +930,9 @@ func (h *WorkspaceHandler) Resume(c *gin.Context) {
 	toResume := []wsInfo{{id, wsName, dbRuntime, tier}}
 	rows, _ := db.DB.QueryContext(ctx,
 		`WITH RECURSIVE descendants AS (
-			SELECT id, name, tier, COALESCE(runtime, 'langgraph') AS runtime FROM workspaces WHERE parent_id = $1 AND status = 'paused'
+			SELECT id, name, tier, COALESCE(runtime, 'claude-code') AS runtime FROM workspaces WHERE parent_id = $1 AND status = 'paused'
 			UNION ALL
-			SELECT w.id, w.name, w.tier, COALESCE(w.runtime, 'langgraph') FROM workspaces w JOIN descendants d ON w.parent_id = d.id WHERE w.status = 'paused'
+			SELECT w.id, w.name, w.tier, COALESCE(w.runtime, 'claude-code') FROM workspaces w JOIN descendants d ON w.parent_id = d.id WHERE w.status = 'paused'
 		) SELECT id, name, tier, runtime FROM descendants`, id)
 	if rows != nil {
 		defer rows.Close()

workspace-server/internal/handlers/workspace_test.go

@@ -35,7 +35,7 @@ func TestWorkspaceGet_Success(t *testing.T) {
 		WithArgs("cccccccc-0001-0000-0000-000000000000").
 		WillReturnRows(sqlmock.NewRows(columns).
 			AddRow("cccccccc-0001-0000-0000-000000000000", "My Agent", "worker", 1, "online", []byte(`{"name":"test"}`),
-				"http://localhost:8001", nil, 2, 1, 0.05, "", 3600, "working", "langgraph",
+				"http://localhost:8001", nil, 2, 1, 0.05, "", 3600, "working", "claude-code",
 				"", 10.0, 20.0, false,
 				nil, 0, false, true, []byte(`{}`)))
 
@@ -60,7 +60,7 @@ func TestWorkspaceGet_Success(t *testing.T) {
 	if resp["status"] != "online" {
 		t.Errorf("expected status 'online', got %v", resp["status"])
 	}
-	if resp["runtime"] != "langgraph" {
+	if resp["runtime"] != "claude-code" {
 		t.Errorf("expected runtime 'langgraph', got %v", resp["runtime"])
 	}
 	// current_task is stripped from public GET response (#955)
@@ -125,7 +125,7 @@ func TestWorkspaceGet_RemovedReturns410(t *testing.T) {
 		WithArgs(id).
 		WillReturnRows(sqlmock.NewRows(columns).
 			AddRow(id, "Old Agent", "worker", 1, string(models.StatusRemoved), []byte(`null`),
-				"", nil, 0, 1, 0.0, "", 0, "", "langgraph",
+				"", nil, 0, 1, 0.0, "", 0, "", "claude-code",
 				"", 0.0, 0.0, false,
 				nil, 0, false, true, []byte(`{}`)))
 	mock.ExpectQuery(`SELECT updated_at FROM workspaces`).
@@ -189,7 +189,7 @@ func TestWorkspaceGet_RemovedReturns410WithNullRemovedAtOnTimestampFetchFailure(
 		WithArgs(id).
 		WillReturnRows(sqlmock.NewRows(columns).
 			AddRow(id, "Vanished", "worker", 1, string(models.StatusRemoved), []byte(`null`),
-				"", nil, 0, 1, 0.0, "", 0, "", "langgraph",
+				"", nil, 0, 1, 0.0, "", 0, "", "claude-code",
 				"", 0.0, 0.0, false,
 				nil, 0, false, true, []byte(`{}`)))
 	// Simulate the row vanishing between the two queries.
@@ -252,7 +252,7 @@ func TestWorkspaceGet_RemovedWithIncludeQueryReturns200(t *testing.T) {
 		WithArgs(id).
 		WillReturnRows(sqlmock.NewRows(columns).
 			AddRow(id, "Audit Agent", "worker", 1, string(models.StatusRemoved), []byte(`null`),
-				"", nil, 0, 1, 0.0, "", 0, "", "langgraph",
+				"", nil, 0, 1, 0.0, "", 0, "", "claude-code",
 				"", 0.0, 0.0, false,
 				nil, 0, false, true, []byte(`{}`)))
 	// last_outbound_at follow-up query (existing path)
@@ -342,7 +342,7 @@ func TestWorkspaceCreate_DBInsertError(t *testing.T) {
 	// Transaction begins, workspace INSERT fails, transaction is rolled back.
 	mock.ExpectBegin()
 	mock.ExpectExec("INSERT INTO workspaces").
-		WithArgs(sqlmock.AnyArg(), "Failing Agent", nil, 3, "langgraph", (*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
+		WithArgs(sqlmock.AnyArg(), "Failing Agent", nil, 3, "claude-code", (*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
 		WillReturnError(sql.ErrConnDone)
 	mock.ExpectRollback()
 
@@ -373,11 +373,13 @@ func TestWorkspaceCreate_DefaultsApplied(t *testing.T) {
 	// Transaction wraps the workspace INSERT (no secrets in this request).
 	mock.ExpectBegin()
 	// Expect workspace INSERT with defaulted tier=3 (Privileged — the
-	// handler default in workspace.go), runtime="langgraph"
+	// handler default in workspace.go), runtime="claude-code"
 	mock.ExpectExec("INSERT INTO workspaces").
-		WithArgs(sqlmock.AnyArg(), "Default Agent", nil, 3, "langgraph", (*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
+		WithArgs(sqlmock.AnyArg(), "Default Agent", nil, 3, "claude-code", (*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
 		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectCommit()
+	mock.ExpectExec("INSERT INTO workspace_secrets").
+		WillReturnResult(sqlmock.NewResult(0, 1))
 
 	// Expect canvas_layouts INSERT (x=0, y=0 — defaults)
 	mock.ExpectExec("INSERT INTO canvas_layouts").
@@ -1429,7 +1431,7 @@ func TestWorkspaceGet_FinancialFieldsStripped(t *testing.T) {
 		WithArgs("cccccccc-0010-0000-0000-000000000000").
 		WillReturnRows(sqlmock.NewRows(columns).
 			AddRow("cccccccc-0010-0000-0000-000000000000", "Finance Test", "worker", 1, "online", []byte(`{}`),
-				"http://localhost:9001", nil, 0, 1, 0.0, "", 0, "", "langgraph",
+				"http://localhost:9001", nil, 0, 1, 0.0, "", 0, "", "claude-code",
 				"", 0.0, 0.0, false,
 				int64(50000), int64(12500), false, true, []byte(`{}`))) // budget_limit=500 USD, spend=125 USD
 
@@ -1489,7 +1491,7 @@ func TestWorkspaceGet_SensitiveFieldsStripped(t *testing.T) {
 				"panic: internal error at /secret/path.go:42",
 				100,
 				"Analyzing customer PII for the Q4 report",
-				"langgraph",
+				"claude-code",
 				"/home/user/secret-projects/client-work",
 				0.0, 0.0, false,
 				nil, 0, false, true, []byte(`{}`)))
@@ -1679,7 +1681,7 @@ func TestWorkspaceCreate_TemplateDefaultsLegacyTopLevelModel(t *testing.T) {
 	}
 	cfg := []byte(`name: Legacy Agent
 tier: 1
-runtime: langgraph
+runtime: hermes
 model: anthropic:claude-sonnet-4-5
 `)
 	if err := os.WriteFile(filepath.Join(templateDir, "config.yaml"), cfg, 0o644); err != nil {
@@ -1695,10 +1697,12 @@ model: anthropic:claude-sonnet-4-5
 	// this assertion should flip back to 1.
 	mock.ExpectExec("INSERT INTO workspaces").
 		WithArgs(
-			sqlmock.AnyArg(), "Legacy Agent", nil, 3, "langgraph",
+			sqlmock.AnyArg(), "Legacy Agent", nil, 3, "hermes",
 			(*string)(nil), nil, "none", (*int64)(nil), models.DefaultMaxConcurrentTasks, "push").
 		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectCommit()
+	mock.ExpectExec("INSERT INTO workspace_secrets").
+		WillReturnResult(sqlmock.NewResult(0, 1))
 	mock.ExpectExec("INSERT INTO canvas_layouts").
 		WithArgs(sqlmock.AnyArg(), float64(0), float64(0)).
 		WillReturnResult(sqlmock.NewResult(0, 1))
@@ -1879,8 +1883,8 @@ func TestWorkspaceCreate_188_NoTemplateNoRuntime_NowMODEL_REQUIRED(t *testing.T)
 	if !bytes.Contains(w.Body.Bytes(), []byte(`"code":"MODEL_REQUIRED"`)) {
 		t.Errorf("bare-body create: expected code=MODEL_REQUIRED in body, got %s", w.Body.String())
 	}
-	if !bytes.Contains(w.Body.Bytes(), []byte(`"runtime":"langgraph"`)) {
-		t.Errorf("bare-body create: expected runtime=\"langgraph\" in 422 body (the gate runs AFTER the langgraph-default assignment so the diagnostic surfaces what runtime WOULD have been used), got %s", w.Body.String())
+	if !bytes.Contains(w.Body.Bytes(), []byte(`"runtime":"claude-code"`)) {
+		t.Errorf("bare-body create: expected runtime=\"claude-code\" in 422 body (the gate runs AFTER the default assignment so the diagnostic surfaces what runtime WOULD have been used), got %s", w.Body.String())
 	}
 }
 

workspace-server/internal/provisioner/localbuild.go

@@ -392,7 +392,7 @@ func remoteHeadShaProd(ctx context.Context, opts *LocalBuildOptions, runtime str
 	}
 	defer func() { _ = resp.Body.Close() }()
 	if resp.StatusCode == http.StatusNotFound {
-		return "", fmt.Errorf("repo not found at %s — runtime %q may not be mirrored to Gitea (only claude-code/hermes/langgraph/autogen today)", apiURL, runtime)
+		return "", fmt.Errorf("repo not found at %s — runtime %q may not be mirrored to Gitea (expected one of claude-code, codex, hermes, openclaw)", apiURL, runtime)
 	}
 	if resp.StatusCode == http.StatusUnauthorized || resp.StatusCode == http.StatusForbidden {
 		return "", fmt.Errorf("auth failure (%d) at %s — verify MOLECULE_GITEA_TOKEN if private repo", resp.StatusCode, apiURL)

workspace-server/internal/provisioner/localbuild_test.go

@@ -161,7 +161,7 @@ func TestEnsureLocalImage_GiteaUnreachable(t *testing.T) {
 	opts.remoteHeadSha = func(ctx context.Context, opts *LocalBuildOptions, runtime string) (string, error) {
 		return "", errors.New("dial tcp: no such host")
 	}
-	_, err := ensureLocalImageWithOpts(context.Background(), "langgraph", opts)
+	_, err := ensureLocalImageWithOpts(context.Background(), "hermes", opts)
 	if err == nil {
 		t.Fatalf("expected error, got nil")
 	}
@@ -259,7 +259,7 @@ func TestEnsureLocalImage_BuildFailure(t *testing.T) {
 	opts.dockerBuild = func(ctx context.Context, opts *LocalBuildOptions, contextDir, tag string) error {
 		return errors.New("Dockerfile syntax error")
 	}
-	_, err := ensureLocalImageWithOpts(context.Background(), "autogen", opts)
+	_, err := ensureLocalImageWithOpts(context.Background(), "codex", opts)
 	if err == nil {
 		t.Fatalf("expected error, got nil")
 	}
@@ -313,7 +313,7 @@ func TestEnsureLocalImage_ConcurrentSameRuntime(t *testing.T) {
 	for i := 0; i < N; i++ {
 		go func() {
 			defer wg.Done()
-			_, _ = ensureLocalImageWithOpts(context.Background(), "langgraph", opts)
+			_, _ = ensureLocalImageWithOpts(context.Background(), "hermes", opts)
 		}()
 	}
 	wg.Wait()

workspace-server/internal/provisioner/provisioner.go

@@ -38,8 +38,8 @@ var ErrNoBackend = errors.New("provisioner: no backend configured (zero-valued r
 // ErrUnresolvableRuntime is returned by selectImage when a workspace
 // names a runtime that has no resolvable image (not in RuntimeImages and
 // no operator-pinned cfg.Image). RFC internal#483 + security review 4269:
-// previously such a request silently fell through to DefaultImage
-// (langgraph) — a user asking for crewai would get a langgraph container
+// previously such a request silently fell through to DefaultImage — a user
+// asking for a removed runtime would get a different container
 // with no signal. The CTO standing directive
 // (feedback_platform_must_hardgate_base_contract) is fail-closed: a
 // named-but-unresolvable runtime must reject with a structured,
@@ -68,8 +68,7 @@ var ErrUnresolvableRuntime = errors.New("provisioner: requested runtime has no r
 // short-circuit pulls entirely if needed.
 var RuntimeImages = computeRuntimeImages()
 
-// DefaultImage is the fallback workspace Docker image (langgraph is the
-// most common runtime). Computed via RegistryPrefix() so the prefix
+// DefaultImage is the fallback workspace Docker image. Computed via RegistryPrefix() so the prefix
 // override applies to the fallback path too.
 //
 // NOTE: Every runtime MUST have an entry in knownRuntimes (registry.go).
@@ -97,7 +96,7 @@ type WorkspaceConfig struct {
 	PluginsPath        string            // Host path to plugins directory (mounted at /plugins)
 	WorkspacePath      string            // Host path to bind-mount as /workspace (if empty, uses Docker named volume)
 	Tier               int
-	Runtime            string // "langgraph" (default) or "claude-code", "codex", "ollama", "custom"
+	Runtime            string // "claude-code" (default), "codex", "hermes", "openclaw", etc.
 	InstanceType       string // Optional CP EC2 instance type override (SaaS only)
 	DiskGB             int32  // Optional CP root volume size override in GiB (SaaS only)
 	Display            WorkspaceDisplayConfig
@@ -139,9 +138,8 @@ type WorkspaceDisplayConfig struct {
 // feedback_platform_must_hardgate_base_contract): if the workspace NAMES a
 // runtime that resolves to no image (not in RuntimeImages, no pinned
 // cfg.Image), reject with ErrUnresolvableRuntime instead of silently
-// substituting DefaultImage. Pre-fix, removing crewai/deepagents/gemini-cli
-// from the catalog left those create requests silently provisioning a
-// langgraph container — the user asked for crewai and got langgraph with no
+// substituting DefaultImage. Pre-fix, removing a runtime from the catalog left
+// those create requests silently provisioning a fallback container with no
 // signal. The error propagates through Start → markProvisionFailed, which
 // already broadcasts WorkspaceProvisionFailed and records the message.
 //
@@ -707,7 +705,7 @@ func buildContainerEnv(cfg WorkspaceConfig) []string {
 		// /app and set ENV ADAPTER_MODULE=adapter, but molecule-runtime is a
 		// pip console_script entry point so cwd isn't on sys.path automatically.
 		// Setting PYTHONPATH from the provisioner fixes every adapter image
-		// (claude-code, hermes, langgraph, …) without needing to PR each
+		// (claude-code, codex, hermes, openclaw, …) without needing to PR each
 		// standalone template repo. Per-template ENV in the Dockerfile can
 		// still override (Dockerfile ENV is overridden by docker -e at runtime).
 		"PYTHONPATH=/app",

workspace-server/internal/provisioner/registry.go

@@ -18,18 +18,15 @@ const defaultRegistryPrefix = "ghcr.io/molecule-ai"
 //
 // Order matters for deterministic test snapshots; keep alphabetical.
 var knownRuntimes = []string{
-	"autogen",
 	"claude-code",
 	"codex",
 	"hermes",
-	"langgraph",
 	"openclaw",
 }
 
 // defaultRuntime is the fallback when a workspace's config doesn't specify a
-// runtime. Picked because LangGraph is the most common in our org templates
-// and has the smallest "first impression" cold-start surface.
-const defaultRuntime = "langgraph"
+// runtime.
+const defaultRuntime = "claude-code"
 
 // RegistryPrefix returns the registry prefix all workspace-template image
 // references should use. Defaults to ghcr.io/molecule-ai (the upstream OSS
@@ -119,4 +116,3 @@ func computeRuntimeImages() map[string]string {
 	}
 	return out
 }
-

workspace-server/internal/provisioner/registry_test.go

@@ -53,8 +53,8 @@ func TestRuntimeImage_AllKnownRuntimes(t *testing.T) {
 		}
 	}
 	// Pin the count so adding a runtime requires explicit test acknowledgement.
-	if len(knownRuntimes) != 6 {
-		t.Errorf("knownRuntimes length = %d, want 6 (autogen, claude-code, codex, hermes, langgraph, openclaw)", len(knownRuntimes))
+	if len(knownRuntimes) != 4 {
+		t.Errorf("knownRuntimes length = %d, want 4 (claude-code, codex, hermes, openclaw)", len(knownRuntimes))
 	}
 }
 

workspace-server/internal/registry/healthsweep.go

@@ -37,7 +37,7 @@ func remoteStaleAfter() time.Duration {
 }
 
 // StartHealthSweep periodically checks all "online" workspaces. For
-// container-backed runtimes (langgraph, claude-code, …) it calls the
+// container-backed runtimes (claude-code, codex, hermes, openclaw) it calls the
 // Docker API via `checker.IsRunning`. For `runtime='external'` (remote
 // agents per Phase 30) it checks heartbeat freshness: a heartbeat older
 // than `REMOTE_LIVENESS_STALE_AFTER` (default 90s) marks the workspace
@@ -79,7 +79,7 @@ func sweepOnlineWorkspaces(ctx context.Context, checker ContainerChecker, onOffl
 	// auto-restart would loop forever (provisioner has no template
 	// for either runtime).
 	rows, err := db.DB.QueryContext(ctx,
-		`SELECT id FROM workspaces WHERE status IN ('online', 'degraded') AND COALESCE(runtime, 'langgraph') NOT IN ('external', 'mock')`)
+		`SELECT id FROM workspaces WHERE status IN ('online', 'degraded') AND COALESCE(runtime, 'claude-code') NOT IN ('external', 'mock')`)
 	if err != nil {
 		log.Printf("Health sweep: query error: %v", err)
 		return
@@ -146,7 +146,7 @@ func sweepStaleRemoteWorkspaces(ctx context.Context, onOffline OfflineHandler) {
 	rows, err := db.DB.QueryContext(ctx, `
 		SELECT id FROM workspaces
 		WHERE status IN ('online', 'degraded')
-		  AND COALESCE(runtime, 'langgraph') = 'external'
+		  AND COALESCE(runtime, 'claude-code') = 'external'
 		  AND COALESCE(last_heartbeat_at, updated_at) < now() - ($1 || ' seconds')::interval
 	`, staleAfterSec)
 	if err != nil {

workspace-server/internal/registry/healthsweep_test.go

@@ -163,7 +163,7 @@ func TestSweepStaleRemoteWorkspaces_MarksStaleAwaitingAgent(t *testing.T) {
 	setupTestRedis(t)
 
 	// Two stale remote workspaces returned by the query
-	mock.ExpectQuery(`FROM workspaces\s+WHERE status IN \('online', 'degraded'\)\s+AND COALESCE\(runtime, 'langgraph'\) = 'external'\s+AND COALESCE\(last_heartbeat_at, updated_at\) < now\(\) - `).
+	mock.ExpectQuery(`FROM workspaces\s+WHERE status IN \('online', 'degraded'\)\s+AND COALESCE\(runtime, 'claude-code'\) = 'external'\s+AND COALESCE\(last_heartbeat_at, updated_at\) < now\(\) - `).
 		WillReturnRows(sqlmock.NewRows([]string{"id"}).
 			AddRow("ws-stale-1").
 			AddRow("ws-stale-2"))
@@ -193,7 +193,7 @@ func TestSweepStaleRemoteWorkspaces_NoStaleWorkspaces(t *testing.T) {
 	mock := setupTestDB(t)
 	setupTestRedis(t)
 
-	mock.ExpectQuery(`FROM workspaces\s+WHERE status IN \('online', 'degraded'\)\s+AND COALESCE\(runtime, 'langgraph'\) = 'external'`).
+	mock.ExpectQuery(`FROM workspaces\s+WHERE status IN \('online', 'degraded'\)\s+AND COALESCE\(runtime, 'claude-code'\) = 'external'`).
 		WillReturnRows(sqlmock.NewRows([]string{"id"}))
 
 	called := 0
@@ -277,7 +277,7 @@ func TestStartHealthSweep_NilCheckerRunsRemoteSweep(t *testing.T) {
 
 	// The goroutine will tick once every 50ms; we give it 200ms then
 	// cancel. sqlmock will satisfy any number of calls.
-	mock.ExpectQuery(`FROM workspaces\s+WHERE status IN \('online', 'degraded'\)\s+AND COALESCE\(runtime, 'langgraph'\) = 'external'`).
+	mock.ExpectQuery(`FROM workspaces\s+WHERE status IN \('online', 'degraded'\)\s+AND COALESCE\(runtime, 'claude-code'\) = 'external'`).
 		WillReturnRows(sqlmock.NewRows([]string{"id"}))
 
 	ctx, cancel := context.WithCancel(context.Background())

workspace-server/internal/registry/hibernation.go

@@ -71,7 +71,7 @@ func hibernateIdleWorkspaces(ctx context.Context, onHibernate HibernateHandler)
 		  AND hibernation_idle_minutes > 0
 		  AND status IN ('online', 'degraded')
 		  AND active_tasks = 0
-		  AND COALESCE(runtime, 'langgraph') != 'external'
+		  AND COALESCE(runtime, 'claude-code') != 'external'
 		  AND last_heartbeat_at IS NOT NULL
 		  AND last_heartbeat_at < now() - (hibernation_idle_minutes * INTERVAL '1 minute')
 	`)

workspace-server/internal/registry/provisiontimeout.go

@@ -21,8 +21,8 @@ type ProvisionTimeoutEmitter interface {
 
 // DefaultProvisioningTimeout is how long a workspace may sit in
 // status='provisioning' before the sweeper flips it to 'failed'.
-// Default for non-hermes runtimes (claude-code, langgraph, crewai,
-// autogen, etc.) which cold-boot in <5 min. The container-launch path
+// Default for non-hermes runtimes (claude-code, codex, openclaw, etc.)
+// which cold-boot in <5 min. The container-launch path
 // has its own 3-minute context timeout (provisioner.ProvisionTimeout)
 // but that only bounds the docker API call — a container that started
 // but crashes before /registry/register never triggers that path and

workspace-server/internal/registry/provisiontimeout_test.go

@@ -336,10 +336,9 @@ func TestProvisioningTimeout_RuntimeAware(t *testing.T) {
 		want    time.Duration
 	}{
 		{"hermes", HermesProvisioningTimeout},
-		{"langgraph", DefaultProvisioningTimeout},
 		{"claude-code", DefaultProvisioningTimeout},
-		{"crewai", DefaultProvisioningTimeout},
-		{"autogen", DefaultProvisioningTimeout},
+		{"codex", DefaultProvisioningTimeout},
+		{"openclaw", DefaultProvisioningTimeout},
 		{"", DefaultProvisioningTimeout},
 		{"unknown-runtime", DefaultProvisioningTimeout},
 	}
@@ -359,17 +358,17 @@ func TestProvisioningTimeout_RuntimeAware(t *testing.T) {
 //
 // Order pinned:
 //
-//   1. PROVISION_TIMEOUT_SECONDS env beats everything (ops debug).
-//   2. Manifest lookup beats hermes special-case + default.
-//   3. Hermes default applies when lookup returns 0 for hermes.
-//   4. DefaultProvisioningTimeout applies when lookup returns 0 for
-//      anything else.
-//   5. Lookup returning 0 for ANY runtime is "no override" — never
-//      a 0-second timeout (which would kill every workspace instantly).
+//  1. PROVISION_TIMEOUT_SECONDS env beats everything (ops debug).
+//  2. Manifest lookup beats hermes special-case + default.
+//  3. Hermes default applies when lookup returns 0 for hermes.
+//  4. DefaultProvisioningTimeout applies when lookup returns 0 for
+//     anything else.
+//  5. Lookup returning 0 for ANY runtime is "no override" — never
+//     a 0-second timeout (which would kill every workspace instantly).
 func TestProvisioningTimeout_ManifestOverride(t *testing.T) {
 	manifest := map[string]int{
 		"claude-code": 900, // 15 min — what an ops manifest bump would set
-		"langgraph":   1200,
+		"codex":       1200,
 		"hermes":      2400, // 40 min — manifest can override hermes default too
 	}
 	lookup := func(runtime string) int { return manifest[runtime] }
@@ -380,7 +379,7 @@ func TestProvisioningTimeout_ManifestOverride(t *testing.T) {
 		want    time.Duration
 	}{
 		{"manifest override beats default for claude-code", "claude-code", 900 * time.Second},
-		{"manifest override applied for langgraph", "langgraph", 1200 * time.Second},
+		{"manifest override applied for codex", "codex", 1200 * time.Second},
 		{"manifest override beats hermes default", "hermes", 2400 * time.Second},
 		{"unknown runtime + no manifest entry → default", "unknown-runtime", DefaultProvisioningTimeout},
 		{"empty runtime + no manifest entry → default", "", DefaultProvisioningTimeout},

workspace-server/internal/router/router.go

@@ -568,7 +568,7 @@ func Setup(hub *ws.Hub, broadcaster *events.Broadcaster, prov *provisioner.Provi
 		var runtime string
 		err := db.DB.QueryRowContext(
 			context.Background(),
-			`SELECT COALESCE(runtime, 'langgraph') FROM workspaces WHERE id = $1`,
+			`SELECT COALESCE(runtime, 'claude-code') FROM workspaces WHERE id = $1`,
 			workspaceID,
 		).Scan(&runtime)
 		return runtime, err

workspace-server/migrations/011_workspace_runtime.sql

@@ -1,2 +1,2 @@
 ALTER TABLE workspaces
-ADD COLUMN IF NOT EXISTS runtime TEXT DEFAULT 'langgraph';
+ADD COLUMN IF NOT EXISTS runtime TEXT DEFAULT 'claude-code';