Enginehaus

Agents are getting smarter. They still don't know your project.

Not the decisions behind it. Not what was tried and rejected. Not why things are shaped the way they are. Enginehaus builds that institutional knowledge structurally — consumed and contributed to with every task, decision, and artifact. It compounds.

Works with any MCP client — auto-configures Claude Code, Cursor, VS Code, Gemini CLI, and more.

Quick Start

npm install -g enginehaus
cd /path/to/your/project
enginehaus init

init detects your AI tools and configures MCP automatically. Restart your AI tool, then say:

"Get my briefing"

That's it. You're coordinating.

From Source

git clone https://github.com/enginehaus/enginehaus.git
cd enginehaus
npm install && npm run build && npm link

cd /path/to/your/project
enginehaus init

The Core Loop

# Pick up work (creates branch, loads full context)
enginehaus task next

# Capture decisions as you go
enginehaus decision log "Chose SQLite over Postgres" \
  -r "Simpler deployment, sufficient for single-node" \
  -c architecture

# Finish cleanly (checks for uncommitted/unpushed work)
enginehaus task complete <id> -s "What you did and why"

When the next session starts — same agent or different, same tool or different — every decision, every phase, every file change is already loaded. No re-explanation. No conflicting choices.

What You Get

Cross-agent memory. Decisions and context persist across sessions and tools. The second agent doesn't re-discover what the first one decided.

Reliability loops. Uncommitted changes block completion. Unpushed work is flagged. Quality gates enforce what instructions can't.

Works everywhere. enginehaus init auto-detects your tools and configures MCP + workflow hooks. Full hook enforcement for Claude Code, Cursor, VS Code, Gemini CLI, and Cline. MCP auto-config for Claude Desktop, Kiro, LM Studio, and more. HTTP/REST API for anything else.

Decision Archaeology. Every architectural choice, tradeoff, and rationale — searchable, auditable, permanent. Next week or next month, any agent can retrieve the reasoning behind any decision.

Configuration

enginehaus init handles this automatically. For manual setup:

| Tool | MCP Config | Hooks | Auto-configured | |------|-----------|-------|-----------------| | Claude Code | .mcp.json | Full | Yes | | Cursor | .cursor/mcp.json | Full | Yes | | VS Code / Copilot | .vscode/mcp.json | Full | Yes | | Gemini CLI | .gemini/settings.json | Full | Yes | | Cline | .clinerules/hooks/hooks.json | Full | Hooks only | | Claude Desktop | Global config | — | Yes | | Kiro CLI | .kiro/settings/mcp.json | — | Yes | | LM Studio | ~/.lmstudio/mcp.json | — | Yes | | Windsurf | ~/.codeium/windsurf/mcp_config.json | — | Manual | | Any MCP client | Varies | — | enginehaus setup --show-config |

Run enginehaus setup --show-config for any tool's config. Works with anything that speaks MCP or HTTP (enginehaus serve on port 47470).

CLI Reference

enginehaus init                         # Set up current directory
enginehaus briefing                     # Project status with insights
enginehaus task list                    # What needs doing
enginehaus task next                    # Claim next priority
enginehaus task complete <id> -s "..."  # Finish with summary
enginehaus decision log "..." -r "..." -c architecture
enginehaus decision list                # What's been decided
enginehaus doctor                       # Diagnose your setup
enginehaus serve                        # Wheelhaus web console + REST API (experimental)

Development

git clone https://github.com/enginehaus/enginehaus.git
cd enginehaus
npm install && npm run build

npm run watch           # Development mode
npm test                # 1140+ tests

Contributing

Contributions welcome. See CONTRIBUTING.md.

License

MIT — see LICENSE.

Documentation | Quick Start | Power User Guide | enginehaus.dev | Issues