Large codebases are hard to understand and even harder to change safely. Most "AI for code" tools treat source as plain text — leading to hallucinated answers and risky edits. CodeWiki is different. It first parses your project into a typed ontology — a graph of entities (packages, files, classes, interfaces, functions, methods, components, types) connected by relations (contains, imports, extends, implements, calls, uses, dependsOn, …). Every feature is grounded in that graph: - Wiki: Auto-generated documentation pages from real code structure, not guesses - Knowledge Graph: Interactive D3 force graph — see how your code fits together - Semantic Search: TF-IDF local vector store ranks code, entities, and docs by meaning - Ask (RAG Chat): Answers with file:line citations, grounded in the ontology - Code Agent: Runs impact analysis on the graph before proposing edits — precise, cross-file aware It runs fully on your machine. Local TF-IDF search needs no external service. Bring your own LLM key (OpenAI / Anthropic / DeepSeek) or run entirely offline with Ollama. Bonus: Ships an MCP server (5 tools) so editors like Cursor or Claude can query your knowledge base, plus a companion VSCode extension. Built with Electron + React 18 + TypeScript + Vite. Available on Windows, macOS, and Linux.

CodeWiki

Turn any codebase into a living wiki, knowledge graph & chat

Turn any codebase into a living wiki, knowledge graph & chat