Yggtirth8205/code-review-graph
↗ GitHubLocal knowledge graph for Claude Code. Builds a persistent map of your codebase so Claude reads only what matters — 6.8× fewer tokens on reviews and up to 49× on daily coding tasks.
3,932
Stars
345
Forks
21
Watchers
8
Open Issues
Safety Rating A
No hardcoded secrets, malicious code patterns, suspicious dependencies, or prompt injection attempts were identified. The project is a straightforward open-source developer tool with MIT license, standard Python packaging, CI via GitHub Actions, and all data stored locally in SQLite. The README contains no embedded instructions targeting AI analysts.
ℹAI-assisted review, not a professional security audit.
AI Analysis
code-review-graph is a local knowledge graph tool for AI coding assistants (Claude Code, Cursor, Windsurf, etc.) that parses codebases into a structural graph using Tree-sitter, stores it in SQLite, and exposes it via the Model Context Protocol (MCP). It performs blast-radius analysis to identify the minimal set of files relevant to a change, enabling significant token reductions (averaging 8.2× vs naive full-codebase reads). It supports 19 programming languages plus Jupyter notebooks, incremental updates triggered by git hooks or file watches, semantic search via vector embeddings, community detection, execution flow tracing, refactoring previews, and multi-repo cross-search.
Use Cases
- Reducing token consumption when using LLM coding assistants on large codebases
- Performing precise, blast-radius-aware AI code reviews
- Incrementally indexing and querying code structure without re-reading entire repos
- Generating architecture overviews and markdown wikis from code community structure
- Detecting risky changes and test coverage gaps before merging pull requests
- Enabling semantic and full-text search over code entities across multiple repositories
Tags
Project Connections
clawvault
→clawvault provides persistent markdown-native memory for AI agents across sessions, while code-review-graph provides structural code context. Together they address complementary sides of AI agent context management: project knowledge vs. session memory.
skyclaw
→skyclaw is an autonomous AI agent runtime with MCP support; code-review-graph exposes an MCP server interface, making it a natural tool plugin for agents like skyclaw to consume structural code context during coding tasks.
kavach
→kavach monitors and restrains AI agent actions at the filesystem level, while code-review-graph helps AI agents make more precise, targeted file reads. Both address safety and correctness of AI agents operating on local codebases, from complementary angles.