Code Cortex MCP

Code Cortex MCP is a local MCP server that helps AI coding agents understand how a codebase fits together. It indexes functions, classes, call paths and service connections, so agents can answer relationship questions without repeatedly searching individual files. It builds on codebase-memory-mcp.

The server runs on macOS, Linux and Windows as a standalone binary, with no Docker, Node or Python runtime required. Its graph data stays on your machine. It's open source under the MIT license and needs no API key for indexing or search. It works with Claude Code, Codex CLI, Gemini CLI, Cursor, Zed and other coding clients; the server's local storage doesn't determine where those clients run their AI models.

Agents can inspect a symbol's definition and callers, trace paths between functions, and assess which code a Git diff could affect. Architecture queries expose packages, HTTP routes, dependency cycles and code hotspots. Cross-service links cover gRPC, GraphQL, tRPC, Socket.IO and pub-sub channels across repositories.

Tree-sitter handles structural parsing, with type-aware call resolution for Go, C, C++, TypeScript/JavaScript, Java, Kotlin, Rust, Python, PHP and C#. C/C++ analysis also handles definitions split across preprocessor branches. Search combines full-text and structural matching with algorithmic semantic similarity, without an external embedding API.

The graph lives in SQLite. Repeated indexing of unchanged code produces identical graphs, and teams can share compressed snapshots so teammates don't need to rebuild the index.