> ## Documentation index
> Fetch the complete documentation index at: [llms.txt](../llms.txt)
> Use this file to discover all available pages before exploring further.

# Docs for AI agents

Every page in these docs has a Markdown version, there’s an index of all of them, and the whole set comes as one file. Here’s what’s where, and how agents should use Sleuth.

## Machine-readable docs

| What | Where | Use it for |
| --- | --- | --- |
| Index | `/llms.txt` | Every page, its Markdown link, and one line on what it answers. Start here. |
| One page | Add `.md` to a page's URL | A single page as clean Markdown |
| Everything | `/llms-full.txt` | All pages in one file, for search or retrieval |

Every page also has **Copy page**, which copies it as Markdown to paste into a chat.

## Instructions for agents using Sleuth

1. **Use the MCP server to act, and these docs to explain.** Server: `https://mcp.sleuthapi.com/mcp`. [MCP server](mcp-overview.md)
2. **Check onboarding first for new users.** Call `get_onboarding_status` before a new user's first task, and follow its `next_action`.
3. **Show the plan before you approve it.** Research runs use the user's credits. When a person is in the loop, approve on their behalf only if they've told you to. A headless job on an organization key is approved automatically, so set its scope up front.
4. **Resolve every name to an ID.** Use the search tools, and never guess IDs. If a term is ambiguous, ask.
5. **Page through data.** Read datasets with `get_dataset` a page at a time instead of downloading whole exports into your context.
6. **Give links as links.** Download links expire within minutes. Don't say a file was downloaded unless you fetched it.
7. **Keep sources and confidence with values.** When you report a value from Sleuth, say where it came from and how confident Sleuth is. [Confidence scores](confidence.md)
8. **Send account problems to Sleuth.** For billing, access, or errors you can't fix, the user can email support@sleuthinsights.com.

## How these pages are written

- One topic per page. The first paragraph answers the page's main question.
- Exact names: tool names, statuses, and limits match the server.
- Credit facts are on [How credits work](how-credits-work.md) and [Plans and pricing](plans.md).
- Terms are defined in the [Glossary](glossary.md).

## Related

- [Sleuth MCP server](mcp-overview.md): Connect any MCP client or agent to Sleuth: run research, look up the knowledge base, build Explore datasets, and search the research library, with 52 tools.
- [How a research run works](mcp-run-lifecycle.md): Create a project, send the question, wait for the status to change, handle approvals and clarifications, then read the datasets and the answer.
- [Glossary](glossary.md): The words you'll see in Sleuth and in these docs.

*Updated November 9, 2026.*
