zaebee/codegraph-brain

Overview

Builds a local code graph for Python and TypeScript projects and exposes it to AI agents through MCP. Provides impact analysis and call tracing. Uses SQLite for storage and runs deterministically.

Best for

Best for
Developers building AI agents that need codebase awareness for Python and TypeScript projects.

Use cases

• Trace call paths before refactoring
• Assess impact of a proposed code change
• Give an AI agent a queryable code graph via MCP

How to use

Install

uvx --from codegraph-brain cgis ingest ./my-project --output graph.db

Tools exposed

• cgis_ingest
• cgis_overview
• cgis_find_symbol
• cgis_analyze_impact
• cgis_trace_flow
• cgis_get_structure
• cgis_context
• cgis_metrics
• cgis_audit_reachability
• cgis_drift
• cgis_validate

Tested with

Claude Desktop, Claude Code, Cursor

Notes

Builds a local code graph for Python and TypeScript projects and exposes it to AI agents through MCP. Provides impact analysis and call tracing. Uses SQLite for storage and runs deterministically.

2 stars on GitHub. Last updated 2026-09-18. Licensed MIT.

Use cases

• Trace call paths before refactoring
• Assess impact of a proposed code change
• Give an AI agent a queryable code graph via MCP

Pros

• Local and deterministic, no cloud dependency
• SQLite storage is simple and portable
• Supports both Python and TypeScript

Cons

• Early-stage project with few stars and limited adoption
• Only covers Python and TypeScript, not other languages
• Requires MCP setup to use with agents

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.