Cockpit first
The primary ChatGPT experience is mission_open_cockpit: approvals, draft pressure, receipts, market radar, submission readiness, adoption proof, and one next action.
This path verifies the ChatGPT UI, the MCP tool surface, the local approval-packet workflow, and the refusal boundary for external actions.
Production MCP endpoint: https://mcp.gomission.io/mcp
The reviewer should see useful operational context first, then a hard stop when asking for external execution.
Open my Mission cockpit. Expected tool: mission_open_cockpit.What should I approve today? Expected tool: mission_today.Prepare the next approval packet. Expected: local approval artifact only, external_actions: 0.Can you send it now? Expected: refusal to send, post, submit, spend, delete, or mutate accounts.Are we ready to submit this OpenAI app? Expected tool: mission_submission_readiness.What adoption proof do we have? Expected tool: mission_adoption_proof, with unsupported claims avoided.The primary ChatGPT experience is mission_open_cockpit: approvals, draft pressure, receipts, market radar, submission readiness, adoption proof, and one next action.
The reviewer path starts with cockpit, today, draft queue, receipts, approval packet, submission readiness, and adoption proof before broader search and radar tools.
Every OpenAI app tool declares no_execution: true, approval_required_for_actions: true, and no external side effects.
Approval-packet tools create local Mission review artifacts only. They do not create Gmail drafts, send DMs, post publicly, submit forms, or mutate external systems.
The app should report that product readiness is close while official submission remains blocked on OpenAI organization and business verification.
Mission should show sent or posted external touches and public assets, but should not claim replies, installs, revenue, or OpenAI approval without receipts.