Your codebase, pre-computed before the LLM sees a single token.
OpenKrak runs the Dorchester engine on your repo — dependency graphs, hotspot scoring, blast radius analysis — and delivers a structured brief directly into Claude’s context before it touches a single file. Built for Claude Code. Supports TypeScript, JavaScript, Python, Go, Rust, Java, and C#.
Requires Node.js ≥ 18.
For Claude Desktop, also add to global config: → MCP → Edit Config File. Or open manually:
Restart Claude Desktop. Look for the hammer icon in the chat input — openkrak tools will appear there.
Open your repo in Claude Code or Claude Desktop and send:
Claude reads CLAUDE.md on every turn — it calls OpenKrak before touching any file.
Add your key to the MCP config to activate unlimited queries:
Full static analysis across your entire repo. Every import, every export, every coupling resolved before the LLM ever runs.
Files ranked by change frequency, coupling density, and cognitive load. The LLM goes where it matters.
Change one file and know exactly what breaks. Dependency chains mapped before you commit.
Execution brief instead of 50,000 tokens of raw code. The LLM reads signal, not noise.
Reset every month.
Saved. Every analysis. Compounding.
Free tier: 15 queries per 24-hour rolling window, no account needed. Pro: unlimited. Activate with OPENKRAK_KEY=your-key in your MCP config.
Secure checkout via Payhip · License key delivered instantly to your email · No subscription, no auto-renewal