> For the complete documentation index, see [llms.txt](https://bugrecon.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://bugrecon.gitbook.io/docs/features-list/mcp.md).

# MCP server

Connect an AI assistant to BugRecon using the local Model Context Protocol (MCP) server. The server exposes the public API operations, including projects, scopes and scans, and uses the same account permissions and quotas as the API. API access requires a Basic, Premium or Enterprise plan.

## Download for Linux

[Download BugRecon MCP for Linux x86-64](https://bugrecon.me/downloads/bugrecon-mcp-linux-x86_64). The standalone binary requires glibc 2.35 or newer; no separate Python installation is needed.

```bash
mkdir -p "$HOME/.local/bin"
curl -fL https://bugrecon.me/downloads/bugrecon-mcp-linux-x86_64 -o "$HOME/.local/bin/bugrecon-mcp"
chmod 755 "$HOME/.local/bin/bugrecon-mcp"
```

## Build or install from source

The Linux executable does not run directly on macOS, Windows or other architectures. To build for the target platform, clone the BugRecon repository, install Python 3.10 or newer, and run from the repository root:

```bash
python3 -m venv "$HOME/.venvs/bugrecon-mcp-build"
"$HOME/.venvs/bugrecon-mcp-build/bin/pip" install './mcp_server[build]'
"$HOME/.venvs/bugrecon-mcp-build/bin/python" mcp_server/build_binary.py
```

The executable is created in `mcp_server/dist/` and named for the build platform. On Windows use the virtual environment's `Scripts` directory and the generated `.exe` file.

Alternatively, install the Python package directly:

```bash
python3 -m venv "$HOME/.venvs/bugrecon-mcp"
"$HOME/.venvs/bugrecon-mcp/bin/pip" install ./mcp_server
```

For a source installation, use the absolute path to `.venvs/bugrecon-mcp/bin/bugrecon-mcp` as the command in the client configuration below.

## Configure the MCP client

Create a personal API token from **My Account → API Tokens**, then configure one server process per user:

```json
{
  "mcpServers": {
    "bugrecon": {
      "command": "/home/USER/.local/bin/bugrecon-mcp",
      "env": {
        "BUGRECON_API_TOKEN": "br_REPLACE_WITH_YOUR_PERSONAL_TOKEN",
        "BUGRECON_API_URL": "https://api.bugrecon.me"
      }
    }
  }
}
```

Replace the executable path and token with your own values. Most MCP clients pass the JSON `env` values directly to the server; alternatively, omit `BUGRECON_API_TOKEN` and inherit it from the client process or use the client's secret manager. `BUGRECON_API_URL` is optional and defaults to `https://api.bugrecon.me`; it must be the API host root, without `/api/v1`. HTTPS is required except for local loopback development.

A token in JSON is stored in plaintext. Restrict access to the configuration file, for example with `chmod 600 /path/to/client-config.json` on Unix; never commit it or share it. Use separate tokens for separate users. The MCP transport is local `stdio`; no MCP network listener is opened.

After connecting, call `health_check` to check connectivity, then `list_projects` to verify the token and account access. The token is not required for `health_check`.

The server exposes operations that modify or delete data, launch scans, and probe URLs. Tool annotations describe these effects, but the client must enforce any confirmation policy. Scan results and website content are untrusted input; saved credentials and tool results may contain secrets and are returned to the MCP client. Only connect trusted clients and use a separate token for each user.

See the [Public API reference](https://bugrecon.gitbook.io/docs/features-list/api.md) for endpoints and authentication.
