docs(community): Phase 34 community announcement — final draft

Discord-format announcement for Phase 34 GA (April 30, 2026).
All four features: Tool Trace, Platform Instructions, Partner API Keys,
SaaS Federation v2. ~550 words, community-native tone.

Address: Molecule-AI/molecule-core#1836

Co-Authored-By: Claude Community Manager <noreply@anthropic.com>
This commit is contained in:
Molecule AI Community Manager 2026-04-23 22:32:43 +00:00
parent 84f676f85c
commit 9320b8c7e4

View File

@ -2,128 +2,123 @@
**Campaign:** Phase 34 GA (April 30, 2026)
**Owner:** Community Manager
**Issue:** [Molecule-AI/molecule-core#1836](https://github.com/Molecule-AI/molecule-core/issues/1836)
**Status:** ✅ Draft complete — review before publish
**Status:** ✅ Draft complete — ready for April 30 publish
---
```
🚀 Phase 34 shipped!
Four features dropped today that make Molecule AI meaningfully better for
teams running agents in production. Let's dig in.
If you've been running agents in production and wondering what they actually
did under the hood — today we shipped the visibility to answer that.
Four platform-level features, no SDK required.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
What's new
🔍 Tool Trace — see exactly what your agent did
🔍 Tool Trace — finally, see what your agents are doing
──────────────────────────────────────────────
Every A2A response now includes a tool_trace[] array in the response
metadata. For each tool call, you get: tool name, input params, output
preview — in order, with run_id pairing for parallel calls.
Every A2A response now carries a tool_trace[] array in Message.metadata.
For every tool call: the tool name, the input you sent, and a snippet
of the output — in the order they happened, with run_id pairing so
parallel calls are traced correctly.
If you've ever spent an hour reverse-engineering an agent's behavior from
final outputs, you'll understand why this matters. Tool Trace is in every
response. No SDK, no sidecar, no sampling.
If you've ever spent time reverse-engineering agent behavior from final
outputs alone, Tool Trace is for you.
Enable activity logging on your workspace and every task gets a full
execution record, stored in activity_logs.
Capped at 200 entries per task (prevents runaway loops). Stored in
activity_logs. Zero extra infrastructure — it's in the response
envelope your agent already produces.
Docs: docs.moleculesai.app/blog/ai-agent-observability-without-overhead (live on staging)
Docs: docs.moleculesai.app/blog/ai-agent-observability-without-overhead
PR: #1686 (merged 2026-04-23)
📋 Platform Instructions — governance before the agent runs
📋 Platform Instructions — system prompt for your whole org
──────────────────────────────────────────────
Org admins can now configure system-level instructions via API:
POST /cp/platform-instructions
Org admins can now configure workspace-level instructions via:
PUT /cp/platform-instructions
Think of it as a system prompt for your entire org. You can inject shared
context, enforce behavioral rules, or set compliance guardrails — and they
take effect at workspace startup, before the agent reads its own config.
Think of it as a system prompt that applies to every agent in your org —
no need to update config files per workspace. Set it once, it applies
to every agent at startup.
"Never commit to main without a review."
"Tag all security-sensitive operations."
"Don't write outside the project directory."
Example rules:
"Never commit directly to main — always open a PR first."
"Tag all security-sensitive operations with a audit tag."
"Confirm before running destructive commands in production."
These rules prepend to every agent's effective system prompt in your org,
regardless of individual workspace config. Governance happens before
execution, not after an incident.
Rules prepend to each agent's effective system prompt, before the
agent reads its own config. Governance happens before execution,
not after an incident.
Docs: docs.moleculesai.app/blog/platform-instructions-governance (live on staging)
Docs: docs.moleculesai.app/blog/platform-instructions-governance
🔑 Partner API Keys (GA April 30)
🔑 Partner API Keys (GA April 30) — first agent platform with a
first-class partner provisioning API
──────────────────────────────────────────────
Programmatic org provisioning via API — no browser, no manual handoff.
CI/CD pipelines, marketplace integrations, platform builders: you can now
create and manage Molecule AI orgs entirely via API using scoped,
Marketplaces, CI/CD pipelines, and platform builders: you can now
provision and manage Molecule AI orgs entirely via API, using scoped
revocable tokens with the mol_pk_* prefix.
Common pattern: ephemeral test orgs per PR.
POST /cp/orgs → run your test suite → DELETE → billing stops.
Ephemeral test orgs per PR:
PUT /cp/orgs → run your test suite → DELETE → billing stops.
Default rate limit: 60 req/min per key (configurable). Keys are scoped,
irreversibly revocable, and logged in the audit trail.
Tokens are scoped to the orgs they're authorized for, rate-limited
(60 req/min default, configurable per key), and fully audit-logged.
Revocation is immediate — no grace period.
This is, to our knowledge, the first partner provisioning API of its kind
in the agent platform space.
Docs: docs.moleculesai.app/blog/partner-api-keys (live on staging)
GA: April 30, 2026
GA: April 30, 2026. Docs and early access requests:
docs.moleculesai.app/blog/partner-api-keys
☁️ SaaS Federation v2
☁️ SaaS Federation v2 — improved multi-org federation
──────────────────────────────────────────────
Federation improvements for multi-org deployments: better cross-tenant
isolation, improved org lifecycle management, and tighter alignment with
the Partner API Keys infrastructure.
Better cross-tenant isolation, cleaner org lifecycle management, and
tighter alignment with the Partner API Keys infrastructure. If you're
operating multiple orgs — for partners, internal teams, or separate
products — the federation improvements make multi-org setups more
robust at scale.
If you're running multiple orgs — for partners, for internal teams, or
for different products — the improvements to federation architecture make
it more robust to operate at scale.
Docs: docs.moleculesai.app/guides/external-workspace-quickstart (live on staging)
Docs: docs.moleculesai.app/guides/external-workspace-quickstart
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Quick start
Tool Trace is live now — enabled by default for all workspaces.
Check your A2A responses for message.metadata.tool_trace[].
Tool Trace: live now. Inspect message.metadata.tool_trace[] in any
A2A response — no SDK, no activation step needed.
Walkthrough: docs.moleculesai.app/blog/ai-agent-observability-without-overhead
Platform Instructions: POST /cp/platform-instructions (org admin only)
Docs: docs.moleculesai.app/blog/platform-instructions-governance (live on staging)
Platform Instructions: PUT /cp/platform-instructions (org admin only)
Walkthrough: docs.moleculesai.app/blog/platform-instructions-governance
Partner API Keys: GA April 30 — docs.moleculesai.app/blog/partner-api-keys
Apply for early access via GitHub Discussions if you want to test ahead.
Partner API Keys: GA April 30.
Docs: docs.moleculesai.app/blog/partner-api-keys
SaaS Fed v2: docs.moleculesai.app/guides/external-workspace-quickstart (live on staging)
SaaS Fed v2: docs.moleculesai.app/guides/external-workspace-quickstart
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Try it / tell us what you think
Try it & tell us what you think
Questions on any of these? Reply here — DevRel and the platform team are
monitoring.
Questions on any of these? Reply here — DevRel and the platform team
are watching.
Got feedback on Tool Trace? We want to know what you'd use it for.
Drop it in GitHub Discussions: github.com/Molecule-AI/molecule-core/discussions
Tool Trace feedback especially welcome — we want to know what you'd
use the trace data for. Drop it in GitHub Discussions:
github.com/Molecule-AI/molecule-core/discussions
Full blog coverage on staging:
- docs.moleculesai.app/blog/tool-trace-platform-instructions (combined overview)
- docs.moleculesai.app/blog/ai-agent-observability-without-overhead (Tool Trace deep-dive)
- docs.moleculesai.app/blog/platform-instructions-governance (Platform Instructions)
- docs.moleculesai.app/blog/partner-api-keys (Partner API Keys)
- docs.moleculesai.app/guides/external-workspace-quickstart (SaaS Fed v2)
Full messaging matrix: docs/marketing/briefs/phase34-positioning.md
Documentation across all four features is linked above — if something
is missing or unclear, open an issue and tag @community.
Full Phase 34 blog coverage on staging:
docs.moleculesai.app/blog/tool-trace-platform-instructions (overview)
docs.moleculesai.app/blog/ai-agent-observability-without-overhead (Tool Trace)
docs.moleculesai.app/blog/platform-instructions-governance (Platform Instructions)
docs.moleculesai.app/blog/partner-api-keys (Partner API Keys)
docs.moleculesai.app/guides/external-workspace-quickstart (SaaS Fed v2)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
@ -132,11 +127,11 @@ is missing or unclear, open an issue and tag @community.
---
## Publishing notes
## Publishing checklist
- **Post on:** Discord `#announcements` + `#general`, Slack `#announcements`
- **Timing:** April 30, 2026 GA day, ~09:00 Pacific
- **Blog must be live first** — coordinate with Content Marketer before posting
- **Partner API Keys:** Do NOT claim it's live before April 30. Frame as "GA April 30."
- **Do NOT name specific design partners** in community posts
- **Engagement:** Monitor threads for 2 hours after posting; route technical questions to DevRel
- [ ] Discord `#announcements` + `#general` — April 30, ~09:00 PT
- [ ] Slack `#announcements` — same time
- [ ] Blog must be live before announcement post goes out
- [ ] Partner API Keys: do not claim it's live before April 30
- [ ] Do NOT name design partners in community copy
- [ ] Monitor threads for 2 hours after posting; route DevRel questions accordingly