Docsearch Mcp

What is Docsearch Mcp?

A local‑first document search and indexing system that provides hybrid semantic + keyword search across local files (including PDFs) and Confluence pages through the Model Context Protocol (MCP). It is designed for AI assistants like Claude Code/Desktop to access documentation, codebases, and research materials.

How to use Docsearch Mcp?

Install globally with npm install -g docsearch-mcp or use with npx. Add the server to your MCP client configuration with the command npx docsearch-mcp start. Use CLI commands such as docsearch-mcp ingest files to index documents and docsearch-mcp search "query" to search. Docker images are also available for containerized deployment.

Key features of Docsearch Mcp

• Hybrid search combining full‑text and vector similarity.
• Index local files, PDFs, and Confluence spaces.
• AI‑powered image description and search for diagrams.
• Database flexibility: SQLite (local‑first) or PostgreSQL (scalable).
• Real‑time file watching with automatic re‑indexing.
• Multiple output formats: text, JSON, and YAML.

Use cases of Docsearch Mcp

• Provide AI assistants with searchable access to local codebases and documentation.
• Index Confluence knowledge bases and enable semantic search across spaces.
• Search and retrieve information from PDF documents with metadata.
• Enable image search in technical diagrams and charts via AI descriptions.

FAQ from Docsearch Mcp

What are the runtime requirements?

Node.js 18+ and TypeScript 5.6. An OpenAI API key is required for embedding generation; Docker is optional.

Where does the indexed data live?

By default, data is stored locally in a SQLite database (configurable path). PostgreSQL can be used instead for scaling.

What file types are supported?

Code files (.ts, .js, .py, .rs, .java, etc.), documentation (.md, .txt, .rst, .yaml, .json), PDFs, and images (.png, .jpg, .gif, .svg, .webp).

Can Docsearch Mcp work without an internet connection?

Local file indexing and search work offline, but embedding generation and image description require an API call to OpenAI (or a compatible TEI endpoint).

How do I configure Confluence indexing?

Set CONFLUENCE_BASE_URL, CONFLUENCE_EMAIL, CONFLUENCE_API_TOKEN, and CONFLUENCE_SPACES in a .env file. Optional filtering by parent page IDs and title include/exclude patterns is supported.