Dash MCP Server

Dash MCP Server is the official Model Context Protocol server from Kapeli, the maker of Dash, the popular macOS offline API documentation browser. It bridges MCP-compatible AI clients (Claude Desktop, Claude Code, and similar) with a local Dash installation so an agent can query and read documentation that is already installed on the user's machine.

The server exposes four tools: listing installed docsets, searching across docsets and snippets, loading a specific documentation page by URL, and enabling full-text search for individual docsets. Because everything runs locally against the Dash app, no API keys or network calls to a third party are required. The server is written in Python and distributed via uvx directly from the GitHub repository.

Notable for being maintained by the original author of Dash itself (Kapeli), so the integration tracks Dash's behavior accurately. The README marks the project as a work in progress, but it covers the core workflows developers expect when pairing offline docs with an AI coding assistant.

Tools

Prerequisites

• macOS (Dash is macOS-only)
• Dash 8 or newer installed (https://kapeli.com/dash)
• Python 3.11.4 or newer
• uv / uvx package manager

Claude Desktop

Add the following to claude_desktop_config.json:

{
  "mcpServers": {
    "dash-api": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/Kapeli/dash-mcp-server.git",
        "dash-mcp-server"
      ]
    }
  }
}

Claude Code

Register the server with a single command:

claude mcp add dash-api -- uvx --from "git+https://github.com/Kapeli/dash-mcp-server.git" "dash-mcp-server"

No authentication is required. The server talks directly to your local Dash installation.

• Let an AI coding agent answer questions using the exact version of a framework's documentation that the developer has installed offline in Dash.
• Search across multiple language and framework docsets (Python, Swift, React, Django, etc.) without leaving the editor or hitting the network.
• Pull full documentation pages into an agent's context to ground code generation in authoritative API references.
• Enable full-text search on a docset on demand so the agent can find prose explanations, not just symbol names.
• Provide air-gapped or offline documentation lookup for developers working without reliable internet access.

• "List all the docsets I have installed in Dash."
• "Search my Dash docs for URLSession dataTask and load the most relevant page."
• "Enable full-text search on the Django docset, then find how to configure middleware."
• "Use Dash to look up the signature of pandas.DataFrame.merge and explain each argument."
• "Find a Swift snippet in Dash for decoding JSON with Codable and adapt it to my model."

• Maintained by Kapeli, the author of Dash itself, so behavior tracks the app accurately.
• Fully local and offline: no API keys, no network calls, no rate limits.
• Covers the core workflow (list, search, load, enable FTS) with a small, focused tool surface.
• Trivial install via uvx straight from the GitHub repo.

• macOS only, because Dash itself is macOS only.
• Marked as a work in progress in the README; feature set is intentionally minimal.
• Requires Dash 8+ and a working Python 3.11.4+ / uv toolchain on the host.

• DocsetMCP: community MCP server that reads Dash-style docsets directly.
• enhanced-dash-mcp: community fork adding fuzzy search, project detection, and Warp Terminal integration.
• apple-docs-mcp: online MCP server focused on Apple Developer Documentation if you do not need offline access.