code-review-graph

Stop burning tokens. Start reviewing smarter.

code-review-graph builds a structural map of your codebase with Tree-sitter, tracks changes incrementally, and provides precise context to AI assistants via MCP so they read only what matters.

Token reduction: 38x–528x across 6 real repositories (compared to dumping full files into context).

Quick Start

pip install code-review-graph
code-review-graph install   # auto-detects all supported AI tools
code-review-graph build     # parse your codebase

Then ask your AI assistant: Build the code review graph for this project

Supported Platforms

Auto-detects and configures for: Codex, Claude Code, CodeBuddy Code, Cursor, Windsurf, Zed, Continue, OpenCode, Antigravity, Gemini CLI, Qwen, Qoder, Kiro, GitHub Copilot (VS Code), GitHub Copilot CLI

Target one with code-review-graph install --platform codex etc.

MCP Tools (30 total)

Includes tools for: codebase querying, dependency analysis, change detection, semantic search, impact analysis (blast radius), cross-repo search, and 5 workflow prompts (review_changes, architecture_map, debug_issue, onboard_developer, pre_merge_check).

Optional Features

DependencyFeature
[embeddings]Local vector embeddings (sentence-transformers)
[google-embeddings]Google Gemini embeddings
[communities]Community detection (igraph)
[enrichment]Python call-resolution (Jedi)
[wiki]Wiki generation with LLM summaries (ollama)
[all]Everything

Config

Create .code-review-graphignore to exclude tracked paths. Embeddings support OpenAI-compatible endpoints (Azure, vLLM, LiteLLM, LocalAI, Ollama in OpenAI mode).

Filter exposed tools: code-review-graph serve --tools query_graph_tool,semantic_search_nodes_tool,detect_changes_tool

Uninstall

code-review-graph uninstall --dry-run  # preview
code-review-graph uninstall            # interactive
code-review-graph uninstall --yes      # non-interactive