CodeGraph
A cross-language code-intelligence engine that builds a semantic graph of a codebase's functions, classes, imports and call chains, exposed through MCP tools, a VS Code extension, a JetBrains plugin and a persistent memory layer.
CodeGraph parses a codebase with tree-sitter across 38 languages and builds a semantic graph of its functions, classes, imports and call chains, so an AI agent can query structured relationships instead of grepping through files. The graph is exposed through 42 Model Context Protocol tools, a VS Code extension, a JetBrains IDE plugin, and a persistent memory layer.
The MCP server indexes the current working directory automatically once added to a client's configuration. The VS Code extension and JetBrains plugin each offer to download a separate native analysis engine on first use, verify it against a published checksum, and share one installed copy between both editors; once present, the engine starts automatically and registers CodeGraph's tools as callable by Copilot or, in JetBrains, the AI Assistant, alongside code-vision annotations and a graph panel. Pre-configured rule files teach agents such as Claude, Cursor, Windsurf, Codex and Cline to use CodeGraph's tools before falling back to grep or reading many files, and a GitHub Action can post an automatic code-graph analysis comment on every pull request.
It is written in C and distributed under the Apache-2.0 license.