MCP Server¶

Expose Drydock analyzers as tools for AI coding assistants (Claude Code, Cursor, Windsurf, VS Code).

Lets you call all 7 analyzers directly as MCP tool calls instead of copy-pasting terminal output into prompts.

Usage¶

CLI¶

Not used via CLI—the MCP server runs as a daemon. Start it once in your coding assistant configuration, then invoke tools from the assistant.

MCP¶

The MCP server exposes these 7 tools:

Tool	Description
`drydock_context`	One-shot: run ALL analyzers at once. Returns structure, platforms, frameworks, codemap, interfaces, boundaries.
`drydock_codemap`	Map a codebase: files, classes, functions, entry points, language stats.
`drydock_boundaries`	Detect component boundaries and extraction risks (the "where to cut" tool).
`drydock_dependencies`	Analyze external and internal dependencies across languages.
`drydock_interfaces`	Extract public API surfaces from all modules.
`drydock_structure`	Analyze file structure: counts, types, sizes, directory depth.
`drydock_platforms`	Detect platforms and runtimes (Node.js, Python, Rust, Go, .NET, etc.) with versions.

Setup Instructions¶

Claude Code¶

Edit ~/.claude/settings.json:

{
  "mcpServers": {
    "drydock": {
      "command": "python",
      "args": ["/path/to/Drydock/Tools/mcp_server.py"]
    }
  }
}

Restart Claude Code.

VS Code¶

Create or edit .vscode/mcp.json in your workspace:

{
  "servers": {
    "drydock": {
      "command": "python",
      "args": ["/path/to/Drydock/Tools/mcp_server.py"]
    }
  }
}

Requires the MCP extension for VS Code.

Cursor¶

Edit .cursor/mcp.json or cursor_settings.json:

{
  "mcpServers": {
    "drydock": {
      "command": "python",
      "args": ["/path/to/Drydock/Tools/mcp_server.py"]
    }
  }
}

Restart Cursor.

Installation¶

pip install mcp

Example Usage in Claude Code¶

Once configured, you can ask:

Use the MCP tool drydock_context on /path/to/my-project to understand its structure.

Or in Claude Code's tool call interface:

{
  "tool": "drydock_context",
  "arguments": {
    "project_path": "/path/to/my-project"
  }
}

This returns a complete JSON object with: - Project metadata - File structure overview - Detected platforms and frameworks - Code map (classes, functions, entry points) - Public interfaces - Component boundaries with extraction risk

Options¶

Each tool accepts parameters:

Tool	Parameters
`drydock_context`	`project_path` (string), `skip_boundaries` (bool, default false)
`drydock_codemap`	`project_path` (string), `max_files` (int, default 2000)
`drydock_boundaries`	`project_path` (string), `min_cluster` (int, default 2)
`drydock_dependencies`	`project_path` (string)
`drydock_interfaces`	`project_path` (string)
`drydock_structure`	`project_path` (string)
`drydock_platforms`	`project_path` (string)