Wraps static dependency analysis and git history mining into eight MCP tools that catch cascading breakage before your AI agent ships code. The core workflow is simple: your agent edits files freely, then calls gate_build before committing to get a PASS/WARN/FAIL verdict based on SCC condensation, PageRank importance scoring, and historical change coupling. Pro tier adds ai_analyze for semantic impact analysis using your own Gemini, OpenAI, or Anthropic key. Supports TypeScript, Python, Go, Rust, Java, C++, Ruby, and Dart out of the box with zero config. Spins up a localhost dashboard showing the full dependency graph with risk indicators and cascade depth. Best for codebases where foundation files have high fan-in and circular dependencies make manual impact assessment unreliable.
claude mcp add --transport stdio khalomsky-syke uvx syke