Codanna, developed by Bartolli, is an mcp server and CLI tool engineered with Rust, providing local code intelligence and semantic code search. This platform tool is designed to enhance AI coding agents by offering deep codebase understanding. It integrates with various client systems, acting as a pivotal connection point for tooling in IDE integrations. Codanna prioritizes local operations, ensuring that code and queries remain on the user's machine by default, making it a suitable choice for privacy-conscious development environments.

What does Codanna connect to?

Codanna functions as an MCP server and CLI for AI coding agents, establishing connections with clients such as Claude Code, Cursor, Windsurf, Codex, and Gemini. It is also compatible with any other MCP-compatible client. The server supports stdio, HTTP, and HTTPS as transport protocols, allowing for flexible integration into diverse development workflows. Its core purpose is to provide local code intelligence and semantic code search, making it a central hub for tooling and IDE integrations within an AI-assisted coding environment.

Core features and tools

Codanna exposes several callable tools, including codanna mcp find_symbol name:"my_function" for locating specific symbols and codanna mcp analyze_impact symbol_name:"my_function" for understanding symbol dependencies. Its codanna mcp semantic_search_with_context query:"recursively extract function calls" command combines five MCP tools to return symbol identity, signature, docstring, callees, callers, and recursive impact analysis. Additionally, codanna dump exports the full code graph as JSON Lines. Setup and indexing are managed by codanna init and codanna index.

Who is it for?

Codanna is designed for developers who use AI coding agents, especially those working with Claude Code. It also targets developers managing agents or pipes who need fast, privacy-focused code intelligence. The program helps these users gain deep codebase understanding and execute complex code exploration tasks efficiently. It serves as an essential component for integrating advanced code analysis into AI-driven development processes. Skip this if you do not utilize AI coding agents or MCP-compatible clients in your development workflow.

How to get started

No credentials are required for its local-first operation. Opt-in remote OpenAI-compatible embeddings are configured via remote_url in settings.toml or by using CODANNA_EMBED environment variables, with the API key being environment-only.

Use cases

• Perform semantic code search, such as "find authentication logic."
• Track relationships and conduct call graph analysis, for example, identifying functions that call "parse_file" or determining what breaks if a symbol is renamed.
• Provide codebase understanding to AI agents by collapsing complex code exploration into single, correlated responses.

Codanna offers a privacy-focused solution for integrating local code intelligence with AI coding agents and MCP clients. Its ability to perform semantic search, analyze impact, and export code graphs makes it a valuable tool for developers. The modular design, supporting various transport protocols, ensures adaptability across different environments. Start by using "codanna mcp semantic_search_with_context" to query for specific code functionalities within your project.