Home

CodeGraph

CodeGraph turns any folder of code into a persistent, queryable knowledge graph, then lets you (or an AI assistant) query that compact graph instead of re-reading the whole codebase. It extracts symbols and relationships across 30+ languages with tree-sitter, clusters them into communities, surfaces the structurally important pieces, and writes machine-readable graphs, human-readable reports, and interactive visualizations.

It is a single static Rust binary (codegraph) with no runtime, plus an MCP server so a coding assistant can consult the graph before grepping or reading files.

60-second tour

codegraph extract .                      # build codegraph-out/ for the current directory
codegraph query "authentication flow"    # get a relevant subgraph back
codegraph explain parse_config           # show a node and its neighbours
codegraph affected parse_config          # what would changing it break?
codegraph serve                          # expose the graph to an assistant over MCP

See Quickstart for a fuller walk-through and Installation to get the binary.

What CodeGraph gives you

• A canonical graph.json you can query without reading source.
• A Markdown report of god nodes, surprising connections, import cycles, and suggested questions (Analysis and Reports).
• 2D, 3D, and SVG visualizations plus GraphML / Cypher / DOT / Obsidian / wiki exports (Output Formats, Visualizations).
• Graph queries: relevant-subgraph search, shortest path, and reverse impact (Querying).
• An MCP server and one-command assistant integration (MCP Server, Assistant Integration).
• Multi-repo federation with real cross-repo edge resolution (Workspaces and Federation).

Documentation map

Getting started: Installation - Quickstart

Concepts: Architecture - Languages

Using CodeGraph: Commands - Extraction - Querying

• Analysis and Reports - Output Formats - Visualizations

Integrations: MCP Server - Assistant Integration

• Ingestion - Semantic Analysis

Scaling: Workspaces and Federation - Incremental Updates - PR Dashboard

Reference: Configuration - Development

Design principles

• Offline by default. A code-only corpus makes no network calls. Only the opt-in --semantic pass over documents calls an LLM (Semantic Analysis).
• Auditable. Every edge carries a confidence level (EXTRACTED, INFERRED, AMBIGUOUS).
• Deterministic. The same input produces the same graph; ids and community numbers are stable across rebuilds.