code-review-graph
Local code intelligence graph giving AI tools precise review context over MCP.
Local-first code intelligence graph that maps dependencies for precise AI coding tool context.
code-review-graph builds a structural map of your codebase using Tree-sitter and serves compact context over the Model Context Protocol. It suits developers trying to prevent AI coding assistants from repeatedly reading entire repositories.
What it is
It parses your repository into a graph of nodes and edges, tracking functions, classes, imports, and calls. When changes happen, blast-radius analysis and incremental updates re-parse only affected files to feed the assistant a precise slice of code.
Who it's for
It suits developers working with large codebases using AI coding tools like Claude Code, Cursor, or Windsurf. It does not suit teams unwilling to manage local graph databases or run command-line installation hooks.
Key features
MCP and CLI Support
Serves compact context directly to AI tools over the Model Context Protocol.
Blast-Radius Analysis
Traces every caller, dependent, and test affected by a file change.
Incremental Updates
Re-parses only files whose SHA-256 hash changed using hooks and watch mode.
Custom Language Mappings
Add unsupported languages by mapping file extensions and node types in TOML.
CI Risk-Scored Reviews
Runs as a GitHub Action to post risk-scored functions and test gaps on pull requests.
Up and running in 3 steps
Our honest verdict
What's great
- Achieves significant token reduction compared to feeding whole codebases to AI assistants.
- Runs locally without sending your source code to an external service.
- Supports a wide array of built-in programming languages and frameworks.
- Integrates automatically with dozens of popular AI coding platforms via CLI setup.
Where it falls short
- Requires maintaining a local graph database and keeping incremental hooks configured.
- Cold build times scale with repository size, taking around 40 seconds for 3,000 files.
- Framework-specific resolvers require explicit imports and visible stubs to work correctly.
Pricing & license
Free under its open source licence, with no paid hosted option mentioned in the README. The only cost is running it yourself on your local machine or CI runner.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Which AI platforms does it support during installation?
It supports codex, claude-code, cursor, windsurf, zed, continue, opencode, antigravity, gemini-cli, qwen, kiro, qoder, copilot, copilot-cli, codebuddy, and hermes.
How are incremental updates triggered?
Updates are triggered via pre-commit hooks, git hooks, or watch mode by tracking file SHA-256 hashes and import/call edges.
