> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.rightbrain.ai/v-1/docs/interop/ai-agent-docs/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.rightbrain.ai/_mcp/server.

# Docs for AI agents

> Help AI assistants and agents discover Rightbrain documentation and read the contract for an individual task.

Rightbrain publishes a version router and version-specific documentation indexes for AI consumption. It also provides a public Markdown endpoint for each task.

## Documentation endpoints

Use the root router to select a documentation version, then use that version's index to discover pages. Use the task endpoint to read one task's contract.

### Select a documentation version

```text
https://docs.rightbrain.ai/llms.txt
```

The root file lists available documentation versions. The default version currently points to:

```text
https://docs.rightbrain.ai/v-1/llms.txt
```

The version-specific file is the complete index for that version. Its page links already target Markdown resources and should be fetched unchanged.

### A single task

```text
https://app.rightbrain.ai/api/v1/public/{project_key}/llms/task/{task_id}
```

Markdown documentation for one task: what it takes, what it returns, and how to call it.

> **Warning**
>
> `project_key` is the project's **`public_routing_key`**, not its id. You will find it on the project — using the id returns nothing.

The router, version index, and task endpoint are public. The task endpoint returns the variables the task takes, its output shape, and the run and token endpoints needed to call it — not the prompts themselves, and not any run history.

## Machine discovery

Give a coding assistant or autonomous agent the root `llms.txt` URL as the entry point. It should follow the selected version link, choose relevant page URLs from that version's index, and fetch those URLs unchanged. The root file routes versions; it is not the complete page index.

## Per-task documentation

To teach an agent how to call one specific task, fetch that task's Markdown docs. Replace `{project_key}` and `{task_id}` with your values:

**`Python`**

```python title="Python"
import requests

url = "https://app.rightbrain.ai/api/v1/public/{project_key}/llms/task/{task_id}"
task_docs = requests.get(url).text
```

**`TypeScript`**

```typescript title="TypeScript"
const url = "https://app.rightbrain.ai/api/v1/public/{project_key}/llms/task/{task_id}";
const taskDocs = await (await fetch(url)).text();
```

The response describes the task's inputs, output schema, and how to invoke it, formatted for an LLM to act on directly.

## Next steps

#### [Run tasks via API](/docs/api/run-tasks)

Invoke a task once your agent knows its shape.

#### [MCP](/docs/interop/mcp)

Expose tasks as callable tools for MCP clients.

> Build reliable AI agents that run inside your existing tools and workflows. Rightbrain developer documentation.