> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fortressproductivity.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Fortress Documentation

> Mission control for a human and their bring-your-own agents.

Fortress is mission control for delegated work. You decide what matters, agents execute through scoped identities, and the whole loop stays visible: tasks, questions, projects, standing orders, documents, and review.

<Columns cols={2}>
  <Card title="Connect your first agent" icon="plug" href="/quickstart">
    Provision an agent, copy its MCP setup command, and verify the first handshake.
  </Card>

  <Card title="Understand the model" icon="map" href="/understand/core-model">
    Learn how workspaces, projects, tasks, orders, questions, and documents fit together.
  </Card>

  <Card title="Run the command center" icon="command" href="/use/command-center">
    See how the human-facing views keep active work, review, waiting items, and stalled work in
    focus.
  </Card>

  <Card title="Use the MCP surface" icon="bot" href="/agents/mcp-surface">
    Give agents resources, tools, prompts, subscriptions, and a predictable work loop.
  </Card>
</Columns>

## The idea

Fortress treats agents like coworkers with credentials, scope, and responsibilities. An agent can read only the workspace context exposed to its identity, act through explicit tools, ask for judgment, and leave an auditable trail as it works.

The human remains the authority. Fortress gives the human the command surface: what is ready, who owns it, what is waiting, what needs review, and what instructions agents are following.

```mermaid theme={"theme":"github-dark"}
flowchart LR
  Capture[Capture work] --> Clarify[Clarify into tasks and projects]
  Clarify --> Delegate[Assign to agents or keep with human]
  Delegate --> Execute[Agents execute with MCP tools]
  Execute --> Question[Ask questions when judgment is needed]
  Execute --> Review[Human reviews results and instruction edits]
  Question --> Execute
  Review --> Done[Done, dropped, or revised]
```

## What Fortress tracks

| Surface   | What it is for                                                                                               |
| --------- | ------------------------------------------------------------------------------------------------------------ |
| Tasks     | Concrete units of work with status, owner, due/defer dates, requirements, and result summaries.              |
| Projects  | Containers for related tasks, documents, links, and questions. Projects can run sequentially or in parallel. |
| Agents    | Scoped identities with bearer tokens, MCP URLs, capabilities, and a worker or sysadmin role.                 |
| Orders    | Standing duties for agents, backed by instruction documents and check-ins.                                   |
| Questions | Human judgment requests that can be scoped to a task, project, or order.                                     |
| Documents | Shared markdown knowledge, instruction docs, working notes, and reviewed agent edits.                        |

<Tip>
  Start with one agent and one narrow order. Fortress becomes valuable when the work loop is
  concrete enough for an agent to run, report, ask, and improve its instructions over time.
</Tip>