OpenDeepWiki: an AI-driven repository knowledge base that generates structured docs, chat interfaces, and MCP endpoints from codebases

What it solves

OpenDeepWiki transforms raw Git repositories, ZIP archives, or local directories into structured, searchable knowledge bases. It eliminates the need to manually write documentation for complex codebases, automatically generating project overviews, wiki catalogs, and detailed document content from the source code.

How it works

The system ingests repository sources and prepares a workspace to analyze the file tree and context. It then uses LLMs to generate README summaries, wiki catalogs, and document content. This processed knowledge is then served through several channels:

• Public Docs Site: An SEO-friendly Next.js site for browsing documentation.
• Chat Assistant: A built-in chat interface and embedded APIs for querying the repository knowledge.
• MCP Endpoints: Model Context Protocol (MCP) endpoints that allow AI agents to access repository-scoped knowledge.
• Visualizations: Generation of Mermaid mind maps and Graphify artifacts to visualize the project structure.

Who it’s for

• Developers who want to automatically generate and host documentation for their open-source or internal projects.
• AI Agent users who need a way to provide their AI assistants with deep, structured context from a specific codebase.
• Project Managers who want a quick overview of a high-level project structure via mind maps and summaries.

Highlights

• Multi-source ingestion: Supports Git URLs, ZIP files, and local directories.
• AI-driven documentation: Automatically generates summaries, catalogs, and multi-language translations.
• AI Agent integration: Provides MCP endpoints for repository-scoped knowledge retrieval.
• Enterprise-ready: Includes admin consoles for managing users, roles, and AI providers, and supports PostgreSQL for production scale.
• Omnichannel chat: Integrates with Feishu, QQ, WeChat, and Slack.