Speakeasy Knowledge Base MCP Server

This is a local, fully self-contained Model Context Protocol (MCP) server containing the complete documentation and formatting rules for Speakeasy SDKs. It uses an embedded ChromaDB database to provide lightning-fast semantic search directly to your AI agents.

Features

• Zero Configuration: The vector database is bundled directly into the package.
• Semantic Search: Agents can query natural language questions (e.g., "How do I configure retries in TypeScript?") and receive the exact relevant documentation chunks.
• FastMCP: Built using Anthropic's official high-performance Python MCP SDK.

Installation

You can install this package globally using pipx or uv:

uv tool install .
# OR
pipx install .

Usage in MCP Clients (Claude Desktop, Cursor, Opencode)

Add the following to your MCP configuration file (e.g., claude_desktop_config.json or your Cursor MCP settings):

{
  "mcpServers": {
    "speakeasy-kb": {
      "command": "uvx",
      "args": ["speakeasy-mcp"]
    }
  }
}

If you installed it globally via pip, you can simply use:

{
  "mcpServers": {
    "speakeasy-kb": {
      "command": "speakeasy-mcp",
      "args": []
    }
  }
}

Available Tools

• query_speakeasy_kb: Searches the Speakeasy documentation for specific rules on SDK formatting, OpenAPI extensions, and configuration.