Differences from codegraph

[Fugui13]

Is there any difference between this repository and codegraph？

[woss]

colbymchenry/codegraph#175 i asked the same question on codegraph repo

[tirth8205]

Fair question, the two projects overlap a lot in concept: both parse a codebase with tree-sitter, store a structural graph in local SQLite with FTS5 search, update incrementally, and expose it over MCP so agents stop re-reading files.

Where they differ, as far as I can tell from codegraph's README:

Implementation. codegraph has a Rust kernel distributed via npm. code-review-graph is pure Python (3.10+), installed with pip/uv, so it is easy to extend but a Rust core will index large repos faster.

Tool surface. codegraph deliberately concentrates on one rich exploration tool (codegraph_explore) plus a CLI. This project ships 30 narrower MCP tools (registered in code_review_graph/main.py) plus 5 prompt templates, so an agent composes small calls (query_graph_tool, get_impact_radius_tool, etc.).

Scope. As the name says, this one leans toward review rather than general exploration: risk-scored change analysis (changes.py, the detect-changes command), execution-flow detection with criticality scoring (flows.py), community detection and architecture overviews (communities.py), review hints (hints.py), a markdown wiki generator, an eval harness, a GitHub Action for PR comments, and a VS Code extension. codegraph has things this project does not, notably framework-aware route detection and cross-language bridging (Swift/ObjC, React Native).

Both do impact/blast-radius analysis. Optional embeddings for hybrid search exist here (embeddings.py) but are off by default.

I have not benchmarked one against the other, so I will not claim either is faster or more accurate. If your agent mainly explores, try both; if you want review-shaped output (risk scores, affected flows, test-gap checks), that is what this repo optimizes for.