better-code-review-graph

Overview

Better Code Review Graph is an MCP server that parses codebases using Tree-sitter, builds a structural graph of functions, classes, and imports, and provides precise context to AI clients like Claude. It offers tools for graph lifecycle management, code queries, impact analysis, and code review context generation. The server supports both local ONNX embeddings and cloud-based embeddings, with configurable options for embedding models and LLM-generated summaries.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• graph

  Graph lifecycle management: build, update, stats, embed, export, summarize

  Try:Build the code graphTry:Update graph embeddingsTry:Export the graph as GraphML
• query

  Predefined pattern queries: callers_of, callees_of, imports_of, etc.

  Try:Find all callers of this functionTry:Show imports of this module
• search

  Search code entities by name/keyword or semantic similarity

  Try:Find functions related to authenticationTry:Search for database connections
• impact

  Blast radius of changed files based on git diff

  Try:Show impact of recent changesTry:What files are affected by this change?
• large_functions

  Find functions/classes exceeding a line-count threshold

  Try:Find functions larger than 100 linesTry:Show me large classes in the codebase
• review

  Generate token-optimized code review context with structural summary

  Try:Prepare a code review for these changesTry:Generate context for the pull request

Comparable tools

Installation

{
  "mcpServers": {
    "better-code-review-graph": {
      "command": "uv",
      "args": ["run", "better-code-review-graph", "stdio"],
      "env": {
        "CRG_DATABASE_URL": "sqlite:///code-review-graph.db"
      }
    }
  }
}

pip install better-code-review-graph
better-code-review-graph init

docker run -it --rm -v "$(pwd):/app" -e CRG_DATABASE_URL="sqlite:///code-review-graph.db" n24q02m/better-code-review-graph:latest"

FAQ

What file formats can be exported from the code graph?

The graph can be exported in 4 formats: graphml (Gephi/Cytoscape), json-ld, dot (Graphviz), and cypher (Neo4j replay).

How are embeddings computed and stored?

Embeddings can be computed using either local ONNX models (qwen3-embed) or cloud APIs. The embeddings are stored in the graph database for semantic search capabilities.

Compare better-code-review-graph with