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

# MCP quickstart

> Check local files for personal data and credentials before your agent opens them.

DataFog MCP connects DataFog Core to Claude Code through five tools. Scan local
files, then create redacted, masked, or removed copies. Tool responses contain
counts, locations, and copy paths, without file contents or matched values.

## Install and connect

You need **Python 3.10 or later**, [uv](https://docs.astral.sh/uv/getting-started/installation/),
and Claude Code running on the machine that holds your files.

**1. Install DataFog MCP**

```bash theme={"theme":"github-dark"}
uv tool install datafog-mcp
```

**2. Choose the folders DataFog can access**

```bash theme={"theme":"github-dark"}
datafog-mcp roots --edit
```

Replace the initial `~` entry with the folders you want to allow, one per line:

```text theme={"theme":"github-dark"}
~/Downloads/customer-exports
```

Check the active roots:

```bash theme={"theme":"github-dark"}
datafog-mcp roots
```

With no roots file or override, access defaults to your home directory.
Narrow the roots before connecting the agent. An empty roots file refuses
every path. Changes take effect immediately.

**3. Connect Claude Code**

```bash theme={"theme":"github-dark"}
claude mcp add datafog --scope user -- "$(command -v datafog-mcp)"
```

Start a new Claude Code session to use the tools.

## Next steps

* [Scan and transform files](/mcp/scanning-and-transformations)
* [Configure allowed folders and copy destinations](/mcp/configuration)
* [Understand detection coverage and privacy boundaries](/mcp/privacy-and-coverage)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.