Codegraph
Static call graph for TS/JS and Python repos: who calls what, impact, and hotspots
- Pricing
- Open source
- Platforms
- CLI, self-hosted, Linux
- Version
- 1.0.0
- Listing updated
- today
Similar MCP servers
About Codegraph
Codegraph is an open-source MCP server for developer tools. It is available on CLI, self-hosted, Linux, macOS and Windows.
The README describes codegraph as a way to see what a code change touches beyond one hop of IDE find-references. It parses a repository with tree-sitter into a graph of functions, classes and files linked by calls, inheritance and imports, stored in SQLite, and answers questions such as who calls a symbol, which code is affected by changing it, and what sits on a path from an entry point to a database or external call.
Every result comes from static analysis, and each edge links back to a source location. Calls that cannot be resolved, such as dynamic dispatch, ambiguous names or shadowed variables, are listed separately as dynamic rather than guessed. The README says no vector store or code upload is required and that query results are reproducible.
The tool is aimed at developers working in large or inherited TypeScript, JavaScript or Python codebases who need to judge the blast radius of an edit. It ships as a CLI and as an MCP stdio server so coding agents such as Claude Code or Cursor can call the same queries.
Tools this server provides (9)
Taken from the project documentation; the running server may differ.
- who_calls
- Lists callers of a symbol.
- impact
- Ranks code affected by a change, with evidence chains.
- trace
- Shows the path between two symbols through the call graph.
- callees
- Lists what a symbol calls.
- dead_code
- Finds functions with no callers.
- cycles
- Detects import cycles.
- hotspots
- Ranks the most central symbols in the graph.
- search
- Searches symbols in the graph.
- archaeology
- Uses git blame to attribute symbol lines to commits and authors.
What it does
- Caller lookup
- Lists every call site of a function with its source location.
- Risk-ranked impact analysis
- Shows the code affected by a change, ordered by risk, with a call chain for each entry.
- Path tracing
- Finds which functions lie between an entry point and a downstream call such as a database.
- Dynamic call flagging
- Marks unresolvable calls as dynamic instead of returning a possibly wrong target.
- Incremental indexing
- Re-parses only changed files and supports a watch mode for automatic rebuilds.
- Git archaeology
- Uses git blame to show which commit and author introduced each line of a symbol.
Good for
- Estimate what breaks before changing a shared TypeScript function
- Find dead code and import cycles in a Python or JS project
- Give a coding agent call-graph queries through MCP instead of text search
- Locate hotspots in an unfamiliar legacy codebase
Pricing
Open source
MIT licensed and self-hosted.
Codegraph FAQ
Which languages are supported?
The README says TypeScript and JavaScript are fully covered and Python is supported, with barrel-style re-export chains handled only on the TypeScript side.
Does it send code to a cloud service?
The README states it needs no vector database and no code upload, and runs from a local build or Docker image.
How does it expose tools to AI clients?
It runs as an MCP stdio server with nine tools, started with a mcp command pointing at a repository root.
Is Codegraph free?
Yes. Codegraph is open source, so you can use and self-host it for free.
What are the best alternatives to Codegraph?
The closest alternatives to Codegraph are Llmgraph, Code-Graph-RAG, codetopo — code graph for AI agents, Code Crawler and Graphify.
Where this comes from
- Description: written from the project's README and its MCP Registry entry.