code-review-graph
by Tirth Patel
Builds a local Tree-sitter graph of a codebase and serves blast-radius and risk-scored review context to coding agents over MCP, plus a CLI and a GitHub Action for pull requests.
Skills
Blast-Radius Analysis
Traces every caller, dependent, and test a changed file could affect, so the assistant reads those files instead of the whole project.
Risk-Scored Reviews
Maps a diff to affected functions, execution flows, and test gaps with detect_changes, and scores each change by risk for review.
PR Review Action
Runs the analysis as a GitHub Action that posts one sticky risk-scored comment per pull request, with an optional fail-on-risk gate.
Incremental Graph Updates
Keeps the graph current through editor hooks, a git pre-commit hook, and watch mode, re-parsing only files whose hash has changed.
Architecture Overview
Clusters the graph into communities with Leiden and produces an architecture map with coupling warnings, hubs, and bridge nodes.
Related Agents
AgentOps for Apify Builders Bundle
Apify actor bundle for agent builders: normalize run traces for QA, control costs, and guard tool calls with a firewall…
CodeGraph
Gives coding agents a local, auto-synced code knowledge graph of symbols, callers, and routes across 20+ languages, que…
gstack
Adds sprint-style slash-command skills to Claude Code and other agents for planning, review, QA, and shipping, plus a b…
Headroom
Compresses tool outputs, logs, RAG chunks, files, and history locally before they reach the model, as a Python or TypeS…