TEKVO / LABS
fmsg workflow showcase · SIMULATION MODE
WORKSPACE / ENGINEERING / RELEASE OPERATIONS

Release control.
With a conversation trail.

Release Relay explores how fmsg messaging can coordinate a developer, QA reviewer, security reviewer and release coordinator across one traceable conversation.

PROJECTIdentity serviceTekvo engineering · demonstration workspace
RELEASE CANDIDATEv0.8.0-rc.1feature/tenant-invitations → main
REQUIRED GATESQA + Security + HumanApproval applies to one revision only
LIVE / FMSG MCP

Connect to the real messaging layer

Checking backend…

The public review demo below is simulated. This separate operator console calls the real fmsg MCP server through our backend. Live messages do not create automated QA/security verdicts.

Owner setup

In your deployment settings, add two server secrets: FMSG_API_KEY (your fmsg agent sub-address key) and RELAY_OPERATOR_TOKEN (a separate random credential of at least 32 characters). Redeploy to apply them. Do not paste the fmsg key into this page. Operators use only the separate operator token below.

Connect verifies identity with whoami and discovers current tools and their schemas. No message is sent by connecting. Send and reply require explicit confirmation. Tool output can contain private messages: use the operator console on a trusted device.

Public visitors can use the demo. Live inbox and send operations require operator access. Tokens stay in page memory only.

Tool argument schema
Connect to discover the real server schema.
No live connection verified yet.

Messages are sent immediately and cannot be recalled. On a timeout, inspect sent messages before retrying. No automatic send retries or model-generated actions.

VOICE / RELEASE ASSISTANT

Say the request. Hear the outcome.

Browser voice · optional

Describe your change, review the transcript, then run the workflow. You can also listen to the current review summary.

Microphone capture starts only when you click Start speaking and permit access. This app does not save audio. A transcript becomes part of local run history only after you use it and run the workflow.

Voice changes the input and readout only. Reviews and fmsg messaging remain simulated. Approval requires the review button.

01 / RELEASE BRIEF

Put a change through review

Synthetic verdicts for demonstration. No code is examined and no fmsg messages are sent.

Ready for a simulated run.

02 / AGENT CONVERSATION
No thread yet
01DeveloperPrepare handoff
02QACheck behavior
03SecurityCheck boundaries
04CoordinatorCollect verdicts
Start a run to see the message handoffs.

Developer → reviewers → release coordinator → you

Recent release runs

Stored in this browser
Connect the idea to real fmsg agents

fmsg’s hosted MCP server gives each connected agent a sub-address. Connect separate agent sessions, verify their identities with whoami, and replace the illustrative roles below with your actual addresses.

Hosted MCP endpoint: https://mcp.fmsg.ai/mcp

1. Developer: send_message starts the release review thread.
2. QA and Security: get_message / get_thread read the handoff.
3. Each reviewer: reply returns a structured verdict in the thread.
4. Coordinator: get_thread collects the review evidence.
5. Human: explicitly approves the exact reviewed revision.

Suggested application payload (not a built-in fmsg schema):
{
  "workflow_id": "release-001",
  "revision": "FULL_COMMIT_SHA",
  "role": "security",
  "verdict": "PASS | BLOCKED",
  "evidence": ["test run URL"],
  "findings": []
}

For a live implementation, accept verdicts only from the configured reviewer addresses for the same revision. Deduplicate processed message IDs, set review deadlines, and block on missing evidence or findings. Treat message text as input, with no authority to override the coordinator’s rules.

This app is a local simulation of that application logic. It has no live MCP connection or deployment action. The approval button records only a demo decision.

Official fmsg MCP documentation ↗ · Web API ↗