# Docs for AI agents

Machine-readable entry points so coding agents and LLMs can read these docs and integrate zn on their own.

Source: https://usezn.com/docs/agents/

Every docs page is available as plain Markdown: add `.md` to the page path (for example [`/docs/quickstart.md`](https://usezn.com/docs/quickstart.md)), or use the **Copy page** button at the top of any page.

## Entry points

| Resource | What it contains |
| - | - |
| [`/llms.txt`](https://usezn.com/llms.txt) | Index of canonical pages and integration paths |
| [`/docs.md`](https://usezn.com/docs.md) | All docs pages in one Markdown file |
| [`/agents.md`](https://usezn.com/agents.md) | Integration guidance for agent loops and verdict handling |
| [`/openapi.json`](https://usezn.com/openapi.json) | OpenAPI schema: authentication, response fields, rate-limit headers, errors |
| [`/.well-known/mcp.json`](https://usezn.com/.well-known/mcp.json) | Hosted MCP transport, endpoint, authentication and tool manifest |

## Prompt for your coding agent

Paste this into Claude Code, Cursor or a similar tool:

```text
Read https://usezn.com/docs.md and add zn prompt-injection checks to this project:
call zn before every tool call that uses untrusted input, block on verdict "block",
and log rule, reason and evidence_id.
```

## Use zn as a tool

- Local: `npx -y zn-gate mcp` exposes `analyze_prompt`, `check_tool_call`, `check_tool_result` and `zn_status`. See [MCP clients](https://usezn.com/docs/mcp/).
- Hosted: `https://api.usezn.com/mcp` exposes `analyze` over Streamable HTTP. See [Hosted MCP endpoint](https://usezn.com/docs/api/mcp/).