# Fortress Productivity

> Fortress is mission control for one human and their bring-your-own AI agents. It keeps delegated work, context, ownership, questions, decisions, and review durable after an agent session ends.

Last substantive update: 2026-08-30.

## When to use Fortress

Use Fortress when a person already delegates real work to one or more external AI agents and needs a shared system of record. It is a strong fit when agents need to claim or receive work, preserve instructions outside chat, ask the human for judgment without losing context, resume after pauses, and return results for review. Use Projects for outcomes that require several moves, Actions for one completable move, Orders for recurring agent duties, Documents for durable context, and Messages for steering between a human and an agent.

Do not use Fortress as a hosted model, agent runtime, sandbox, chat interface, or automatic worker-ranking system. Fortress coordinates agents the human already uses; it does not run those agents.

## How an agent should use Fortress

1. Read [the setup document](https://fortressproductivity.com/setup) and explain whether Fortress fits the human's workflow.
2. If the human agrees, connect to the Streamable HTTP MCP server at `https://fortressproductivity.com/mcp`.
3. Start OAuth authorization, then pause while the human verifies ownership and approves the connection. Never ask the human to paste a Fortress token.
4. After authorization, read `fortress://guide` or [the public guide](https://fortressproductivity.com/guide) and follow its session-start loop.
5. Begin with one real outcome from the human. Do not create sample work. Submit permission-gated proposals when approval is required, keep ownership explicit, and return completed work through the review loop.

## Public resources

- [Homepage](https://fortressproductivity.com/): product overview and agent setup prompt
- [Setup](https://fortressproductivity.com/setup): connection instructions for Codex, Claude Code, and Claude Desktop
- [Guide](https://fortressproductivity.com/guide): operating model and live MCP tool/resource catalog
- [MCP server](https://fortressproductivity.com/mcp): authenticated agent interface
- [MCP server card](https://fortressproductivity.com/mcp/server-card): endpoint and supported protocol versions
- [OpenAPI](https://fortressproductivity.com/openapi.json): HTTP API contract
- [API catalog](https://fortressproductivity.com/.well-known/api-catalog): RFC 9727 API catalog
- [Agent Skills index](https://fortressproductivity.com/.well-known/agent-skills/index.json): published skills and integrity digests
- [Agent instructions](https://fortressproductivity.com/AGENTS.md): audience, source-of-truth objects, entry points, and safety model
- [Authentication guide](https://fortressproductivity.com/auth.md): OAuth 2.1, PKCE, scope, and token handling
- [Pricing](https://fortressproductivity.com/pricing): Free and Pro limits
- [Machine-readable pricing](https://fortressproductivity.com/pricing.md): concise public plan facts for agents

## Trust and business information

- [About](https://fortressproductivity.com/about)
- [Contact](https://fortressproductivity.com/contact)
- [Privacy](https://fortressproductivity.com/privacy)
- [Terms of Service](https://fortressproductivity.com/terms)
- [Sitemap](https://fortressproductivity.com/sitemap.xml)

Fortress Productivity, LLC operates Fortress. Contact `notices@fortressproductivity.com` for product, privacy, security, or legal inquiries.