docs-server

Category: Documentation & Learning Resources
Tags: mcp, documentation, search, integration

Description

A Model Context Protocol (MCP) server that scrapes, processes, indexes, and searches documentation for third-party libraries and packages. It supports versioning and hybrid search and is directly relevant as an MCP server.

Source URL: https://ubos.tech/mcp/docs-mcp-server-2/

Features

• Documentation Fetching & Scraping: Fetches content from specified URLs and scrapes documentation for software libraries and packages.
• Processing & Indexing: Processes and splits documentation into meaningful chunks using semantic splitting techniques.
• Vector Embedding: Generates vector embeddings (default: OpenAI, but supports Google Vertex, Gemini, AWS Bedrock, Microsoft Azure) and stores them in an SQLite database.
• Hybrid Search: Combines vector similarity search (using sqlite-vec) and full-text search (using FTS5) for efficient and accurate hybrid search results.
• Versioning Support: Stores and queries documentation for different library versions, including unversioned content.
• MCP Tools Exposure: Exposes MCP tools for configuration and integration with AI assistants.
• Customizable Embedding Model: Environment variables allow configuration of the embedding provider and model.
• Multiple Deployment Options:
  • Docker: Recommended for ease of use; no need for Node.js.
  • npx: For cases needing local file access, requires Node.js.
  • From Source: For development and advanced setup.
• CLI Tools:
  • Manage documentation (fetch, scrape, search, list, remove, find-version).
  • CLI can be used via Docker or npx, matching the server environment.
• Debugging Tools: MCP Inspector package script for debugging when running from source.
• Automated Release Process: Uses semantic-release and Conventional Commits for changelog, version bumps, npm publish, and GitHub releases.
• Architecture Documentation: Detailed architecture and design principles available in the project documentation.

CLI Commands

• fetch-url: Fetches a single URL and converts its content to Markdown (no crawling or storing).
• scrape: Scrapes and indexes documentation from a URL for a specific library.
• search: Searches the indexed documentation, filterable by version.
• find-version: Finds the best matching version for a library.
• list: Lists all indexed libraries.
• remove: Removes indexed documentation for a specific library and version.

Configuration

• Embedding Model: Configurable via environment variables (DOCS_MCP_EMBEDDING_MODEL).
• Supported Providers: OpenAI (default), Google Vertex, Gemini, AWS Bedrock, Microsoft Azure, and OpenAI-compatible APIs.
• Vector Dimension: Supports embedding models with vector dimension ≤ 1536 (with exceptions for providers like Gemini).
• Environment Variables: Can be set for Docker, npx, or source deployments.

Deployment Options

• Docker: Recommended for most users; environment variables passed via -e.
• npx: For local file system access (requires Node.js).
• From Source: For development, requires cloning the repo, installing dependencies, and configuration.

Pricing

No pricing information is provided. The project appears to be open-source.

More information: docs-server documentation