---
title: MCP
description: Let AI assistants work with your tables through the Anyrow MCP server, and read these docs through the docs server.
url: https://anyrow.ai/docs/mcp
canonical_url: https://anyrow.ai/docs/mcp
last_updated: '2026-10-16'
---

# MCP

Anyrow speaks the [Model Context Protocol](https://modelcontextprotocol.io) twice: the
**Anyrow MCP server** lets an assistant read and extract into your tables with your API key,
and the **docs MCP server** lets it search and read these docs while it writes code.

## Anyrow MCP server

`@anyrow/mcp-server` runs on your machine over stdio and calls the API with an API key. Give it a
key with only the [permissions](/docs/authorization) the assistant needs: `tables:read` to
read, plus `extract:write` to extract.

```bash title="Claude Code"
claude mcp add anyrow \
  --env ANYROW_API_KEY=ak_... \
  --env ANYROW_BASE_URL=https://api.anyrow.ai \
  -- npx -y @anyrow/mcp-server
```

```json title="Claude Desktop"
{
  "mcpServers": {
    "anyrow": {
      "command": "npx",
      "args": ["-y", "@anyrow/mcp-server"],
      "env": {
        "ANYROW_API_KEY": "ak_...",
        "ANYROW_BASE_URL": "https://api.anyrow.ai"
      }
    }
  }
}
```

```json title="Cursor"
{
  "mcpServers": {
    "anyrow": {
      "command": "npx",
      "args": ["-y", "@anyrow/mcp-server"],
      "env": {
        "ANYROW_API_KEY": "ak_...",
        "ANYROW_BASE_URL": "https://api.anyrow.ai"
      }
    }
  }
}
```

```json title="VS Code"
{
  "servers": {
    "anyrow": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@anyrow/mcp-server"],
      "env": {
        "ANYROW_API_KEY": "ak_...",
        "ANYROW_BASE_URL": "https://api.anyrow.ai"
      }
    }
  }
}
```

Claude Desktop reads `claude_desktop_config.json`, Cursor `.cursor/mcp.json` and VS Code
`.vscode/mcp.json`.

| Variable | Holds |
| :-- | :-- |
| `ANYROW_API_KEY` | The project API key it calls with |
| `ANYROW_BASE_URL` | `https://api.anyrow.ai` |

### Tools

| Tool | What it does | Permission |
| :-- | :-- | :-- |
| `anyrow_table_list` | List the project's tables | `tables:read` |
| `anyrow_table_get` | Get one table and its columns | `tables:read` |
| `anyrow_row_list` | List a table's rows, with search, filter and paging | `tables:read` |
| `anyrow_row_get` | Get one row | `tables:read` |
| `anyrow_extract_once` | Extract data from text or a file | `extract:write` |
| `anyrow_suggest_schema` | Suggest a table's columns from a sample | `extract:write` |

Each tool's input is the matching endpoint's: path parameters in `params` (`project_id`,
`table_id`, `row_id`) and query parameters in `search`.

## Docs MCP server

The docs are an MCP server too, hosted at `https://anyrow.ai/docs/mcp` (Streamable HTTP, no
key), so an assistant can search them and read any page or endpoint while it works.

```bash title="Claude Code"
claude mcp add --transport http anyrow-docs https://anyrow.ai/docs/mcp
```

```json title="Cursor"
{
  "mcpServers": {
    "anyrow-docs": { "url": "https://anyrow.ai/docs/mcp" }
  }
}
```

```json title="VS Code"
{
  "servers": {
    "anyrow-docs": { "type": "http", "url": "https://anyrow.ai/docs/mcp" }
  }
}
```

| Tool | What it does |
| :-- | :-- |
| `search_docs` | Search the guides and the API reference; returns pages with their Markdown URLs |
| `get_page` | Get a page as Markdown by its path (`/docs/quickstart`, `/docs/api/rows/list`) |
| `get_operation` | Get one endpoint's reference by operation id (`row.list`) or method and path |
| `list_pages` | List every page, guides and API resources, with its Markdown URL |

The **Copy page** menu on every page has the same command, and one-click links for Cursor and
VS Code. See [Build with AI](/docs/build-with-ai) for the docs as Markdown and `llms.txt`.
