bartolli/codanna
Local code intelligence MCP server and CLI for AI coding agents
What it solves
Codanna provides AI assistants with a structured understanding of a codebase, eliminating the need for basic keyword searches (grep) and repetitive back-and-forth interactions. It allows agents to understand function relationships, trace dependencies, and find implementations based on intent rather than just names.
How it works
Codanna indexes the source code and documentation of a project. It uses semantic search to match natural language queries to code intent and tracks relationships like call graphs and dependencies across files. It integrates with AI assistants via the Model Context Protocol (MCP), supporting various transports like stdio, HTTP, and HTTPS.
Who it’s for
Developers and AI agents (such as Claude Code, Cursor, and Windsurf) who need deep, context-aware navigation and search within large codebases during R&D and pair programming.
Highlights
- Semantic Search: Finds functions and logic by what they do using natural language queries.
- Relationship Tracking: Provides instant call graphs and dependency analysis across files.
- Document Search: Indexes markdown and text files for RAG workflows to query documentation alongside code.
- MCP Protocol Integration: Native compatibility with Claude, Gemini, and Codex via the Model Context Protocol.
- High Performance: Capable of parsing 75,000+ symbols per second with sub-10ms lookups.