CGraph is an open-source code-intelligence engine that turns any repository into a queryable graph of its symbols and relationships — a self-hosted, free-to-fork alternative to the proprietary code-search and analysis tools most teams rent. It runs on your machine, understands 11+ languages via tree-sitter, and answers structural questions (what calls this, what breaks if I change it, what's the shortest path between two functions) in milliseconds.
Why "code intelligence" is usually proprietary
The tools that answer "where is this used across the whole codebase?" — cross-repo search, call graphs, impact analysis — have historically been closed, hosted, and priced per seat. Your source code, and the graph derived from it, lives on someone else's servers. For a lot of teams that's a non-starter: the codebase is the crown jewels, and a per-developer subscription to understand your own code is a hard sell.
CGraph takes the opposite position. The extraction is deterministic and runs locally, the output formats are open, and the whole thing is on GitHub to use, fork, and extend.
What CGraph actually does
- Extract. A one-shot scan of your source tree produces a graph of definitions and references. It's deterministic,
.gitignore-aware, and skips generated directories automatically. - Serve. A daemon keeps the graph warm and watches your tree — edits fold in incrementally within seconds, so queries stay fast without a full rebuild.
- Query. Warm queries return in roughly 10ms: symbol search, impact analysis, shortest paths between symbols, and budgeted context for feeding an LLM.
Because the graph is a real data structure rather than a text index, the answers are structural, not string matches — "everything that transitively depends on this function," not "every file that mentions its name."
Open export, no lock-in
A proprietary tool's insights are trapped in its UI. CGraph exports the same graph six ways — JSON, HTML, SVG, Obsidian, Cypher, and call-flow — so you can drop it into a graph database, browse it in your notes, embed a diagram in a doc, or pipe it into your own tooling. The data is yours.
Built for AI coding agents
The reason code intelligence matters more now than it did five years ago: AI coding agents are only as good as the context you give them. Dumping raw files at a model is wasteful and often wrong. CGraph exposes eight tools over MCP — query, explain, impact, path, context, and more — so agents like Claude Code, Codex, and Cursor can ask the graph precise questions and pull budgeted context: just enough of the repo to answer, and no more.
That's the difference between an agent guessing from a handful of open tabs and one reasoning over the actual structure of your codebase.
Try it
CGraph is open source and self-hosted. Browse the repository, clone it, and build a graph of a repo you already know — the fastest way to see what structural queries buy you. For more on the AI-context angle, see the rest of our writing.