Enginehaus is an MCP server that treats AI collaboration as coordination, not command. It’s the layer that makes context compound — decisions persist, quality is enforced structurally, and every session starts where the last one left off.
The problem it solves
There’s a pattern we call cognitive drift: AI agents lose context between sessions and re-discover what you already decided. You brief Monday’s agent, then brief Wednesday’s agent the same way. Decisions evaporate. Quality varies. The human becomes the bottleneck — not because the tools are slow, but because they forget.
Enginehaus captures decisions, enforces quality gates, and loads context automatically — as a byproduct of working, not a chore you maintain.
Key ideas
- Structure over instruction. Agents forget instructions; they can’t skip code. If it matters, it’s a gate.
- Coordination, not command. AI amplifies your intent, but only if your intent survives between sessions. Enginehaus makes it survive.
- Transparency by default. Every decision is logged with reasoning. Every phase transition is tracked. The audit trail is the product.
Beyond software
Built for software engineers first, Enginehaus now supports domain profiles — configurable phases, decision categories, and quality gates for different kinds of work. Writing profiles track tone and narrative decisions with editorial checkpoints. Research profiles enforce methodology rigor and citation verification. Profiles are JSON files — anyone can create one for their domain.
Technical approach
Built on the Model Context Protocol. Works with Claude Desktop, Claude Code, Cursor, Windsurf, Gemini CLI, Kiro, and any MCP client. Runs locally. Your data stays on your machine.
Enginehaus is the technical foundation for the other Applied Humanism experiments — the coordination layer they build on.
Try it
npx enginehaus init