TempoGraph parses your codebase into a dependency graph using tree-sitter and exposes 24 MCP tools for structural code analysis. The main tool, prepare_context, takes a task description and returns the exact files your agent should read based on dependency analysis rather than text search. Beyond that, you get blast_radius for impact analysis, focus for symbol exploration, hotspots for risk ranking, and dead_code detection. Benchmarked on real PRs from Django, Flask, and FastAPI with 27% average F1 improvement across GPT-4o, GPT-4o-mini, and Qwen models. Handles Python, TypeScript, JavaScript, Rust, Go, Java, C#, Ruby deeply, plus 170+ languages via generic handlers. Results cache in SQLite with content hashing so branch switching stays fast.
claude mcp add --transport stdio elmoaid-tempograph uvx tempograph