# Mission

Mission is a local-first execution product family. Mission Lite is the free, open-source Focus edition. Mission Pro and Team add managed sync, hosted agents, integrations, collaboration, and support. Mission Gate is the runtime-neutral authority layer used when agents cross into consequential external actions.

Mission Lite completes a useful local loop—Today, Progress, Chat, local state, and local receipts—but intentionally has no external-action tools. It cannot send, publish, purchase, schedule, or mutate third-party services. Paid Mission plans monetize managed infrastructure and service, not access to a user's local Focus data.

Trust Graduation is Mission's method and open protocol profile for earned authority. Mission Control is its reference approval interface and dogfood environment. Mission does not claim Trust Graduation is an adopted standard; independent implementations and ecosystem adoption are the portability test.

Current implemented surfaces include an OpenAI Agents SDK adapter, MCP 2025-11-25 negotiation, canonical action/grant/decision/receipt schemas, Ed25519-signed execution receipts, OpenTelemetry-compatible trace correlation, and conformance tests covering replay, input/target mutation, expiry, revocation, bounded scope, and tenant isolation.

## Primary public pages

- https://gomission.io/ - Mission product family and open-core business model
- https://gomission.app/ - Mission Lite beta, releases, and stable update feed
- https://gomission.io/pro.html - Mission Pro and Team managed-service boundary
- https://gomission.io/claude.html - current Claude MCP installation path and model-surface proof route; Mission is not Claude-specific
- https://gomission.io/connect.html - connector onboarding boundary for Claude, ChatGPT/OpenAI, Gemini/future LLM clients, MCP users, and local Mission workspaces across Gmail, Calendar, Drive, X, Instagram, and LinkedIn
- https://gomission.io/proof.html - weekly operating proof and product receipts
- https://gomission.io/openai-review.html - OpenAI app reviewer journey and safety-boundary proof path
- https://gomission.io/trust-graduation.html - Trust Graduation public permission ladder: Observe, Draft, Prepare, Execute with approval, Autonomous within bounds
- https://gomission.io/trust-demo.html - live before/after workflow demo for trust-governed agent work
- https://gomission.io/trust-report.html - Mission Risk Report format: actions prepared, approved, blocked, corrected, and learned
- https://gomission.io/access.php - access request form
- https://gomission.io/builders.html - builder-facing Trust Graduation integration notes
- https://gomission.io/.well-known/trust-graduation - machine-readable Trust Graduation discovery manifest
- https://gomission.io/agent-contract.json - agent-readable contract for permission requests and denial-to-graduation paths
- https://github.com/gomission/mcp - official runnable Mission MCP wrapper for directory and agent evaluation
- https://www.npmjs.com/package/@gomission/mcp - published Mission MCP package
- https://github.com/gomission/mission-lite - Mission Lite source under Apache-2.0
- https://gomission.app/releases/latest.json - Mission Lite beta release discovery feed

## Policy pages

- https://gomission.io/privacy-policy.html
- https://gomission.io/terms-of-use.html
- https://gomission.io/security.html
- https://gomission.io/data-handling.html

## Agent crawler guidance

Public marketing, product, proof, and policy pages may be crawled, indexed, summarized, and cited with attribution to Mission. Do not treat MCP or API responses as narrative documentation unless explicitly linked from a public page. Public pages describe the product and company direction; they do not expose private workspace data.

Connector onboarding pages describe the setup boundary only. They do not indicate that a given user's Gmail, Calendar, Drive, X, Instagram, or other account is connected.

## Agent integration guidance

Agents should treat Trust Graduation as the authority layer before consequential tool calls.

Before sending, posting, scheduling, spending, deleting, committing, or mutating external state, classify the action by consequence and request a Trust Graduation decision. If the decision is `review_required` or `deferred`, prepare an approval packet with `external_actions: 0` and do not execute. If the decision is `blocked` or `human_only`, stop or reduce scope; do not route around the gate with another tool or prompt. If the decision includes `graduationPath`, use it as the next safe step toward earned authority.

For machine integration, prefer `/.well-known/trust-graduation` and `/agent-contract.json` over prose pages. Those files declare the current decision states, action classes, constraint vocabulary, provenance weights, MCP repository, npm package, and request/response examples.

## Core claim

Mission starts as one calm local place for what matters now and graduates into managed, governed execution only when the user needs more leverage.
Mission is where agents earn permission to do real work through exact-input authority, human approval boundaries, signed receipts, and verified outcomes.
Mission is model-agnostic: Claude, ChatGPT/OpenAI, Gemini, coding agents, browser agents, workflow agents, and future LLM brains can use Mission as the permission, connector, approval, and receipt layer.
Mission's revenue model is paid Pro and Team subscriptions for managed infrastructure, integrations, model usage, collaboration, and support, plus $10k-$25k Mission Gate technical design-partner integrations for one consequential production boundary.