How It Works

CodeGraph turns source code into a queryable graph in four stages.

files → Extraction (tree-sitter) → DB (nodes/edges/files)
            ↓
      Resolution (imports, name-matching, framework patterns)
            ↓
      Graph queries (callers, callees, impact)
            ↓
      Context building (markdown / JSON for AI consumption)

1. Extraction

tree-sitter parses source into ASTs. Language-specific queries extract nodes (functions, classes, methods, types…) and edges (calls, imports, extends, implements). Heavy parsing runs off the main thread.

2. Storage

Everything goes into a local SQLite database (.codegraph/codegraph.db) with FTS5 full-text search. CodeGraph uses native better-sqlite3 when available and transparently falls back to a WASM backend; codegraph status shows which is live.

3. Resolution

After extraction, references are resolved: function calls → definitions, imports → source files, class inheritance, and framework-specific patterns. Some dynamic-dispatch boundaries (callbacks, observers, React re-render, JSX children) are bridged by synthesizers so flows connect end-to-end. See Resolution & Frameworks.

4. Auto-sync

The MCP server watches your project using native OS file events (FSEvents / inotify / ReadDirectoryChangesW). Changes are debounced, filtered to source files, and incrementally synced — the graph stays fresh as you code, with no configuration.