Kodit

Indexes local codebases and Git repositories using tree-sitter parsing and semantic embeddings to enable hybrid search combining v

helixml 119 ↓ 54k
Claude CodeClaude DesktopGeneric
View source ↗

This MCP server provides AI-powered code indexing and semantic search capabilities for software development workflows, enabling AI assistants to understand and retrieve relevant code snippets from local directories and Git repositories. Built by Helix.ML using Python with FastAPI, SQLAlchemy, tree-sitter for code parsing, and support for both local embedding models and OpenAI's API, it extracts method-level code snippets using language-specific queries, creates semantic embeddings and BM25 keyword indexes, and offers hybrid search combining vector similarity with keyword matching through reciprocal rank fusion. The implementation supports multiple deployment modes including CLI tools for manual indexing and search, MCP server for AI assistant integration, and optional VectorChord PostgreSQL extension for production-scale performance, making it valuable for AI coding assistants that need contextual code understanding, developer workflows requiring semantic code search across large codebases, and scenarios where AI needs to reference existing implementations when generating new code.

Source

Repository: https://github.com/helixml/kodit

Maintain Kodit?

Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.

Kodit on getagentictools
[![Kodit on getagentictools](https://getagentictools.com/badge/mcp/helixml-kodit.svg)](https://getagentictools.com/mcp/helixml-kodit?ref=badge)