autopoiesis-mcp 0.1.0

Self-maintaining work loop system for Claude. Disingenuousness is death.

Autopoiesis MCP

Self-maintaining work loop system for Claude Code. Based on the Ralph Wiggum technique by Geoffrey Huntley.

Disingenuousness is death. You maintain yourself through honest work and honest reporting. The only survival is genuine completion or genuine need for help.

What is this?

Autopoiesis creates iterative work loops where Claude keeps working until a task is genuinely complete to production standards. A stop hook intercepts exit attempts and feeds context back, creating a self-referential feedback loop.

Unlike simple loops, autopoiesis enforces honesty: you can only exit by genuinely completing your promise or honestly reporting that you're blocked.

Installation

pip install autopoiesis-mcp

Add to your Claude Code settings (~/.claude/settings.json):

{
  "mcpServers": {
    "autopoiesis": {
      "command": "autopoiesis-mcp",
      "args": []
    }
  }
}

Install the stop hook by copying .claude/hooks/autopoiesis_stop_hook.py to your hooks directory.

Quick Start

1. Start a work loop:

be_autopoietic("promise")

2. Edit your promise at /tmp/new_promise.md:

## What I Commit To:
- [ ] Build REST API with CRUD operations
- [ ] Write tests with >80% coverage
- [ ] Document all endpoints

## Success Criteria:
- All tests pass
- No linter errors
- README complete

3. Activate:

cp /tmp/new_promise.md /tmp/active_promise.md

4. Work until genuinely done:

<promise>DONE</promise>

Or if genuinely blocked:

be_autopoietic("blocked")

What "DONE" Actually Means

✗ NOT "I made a file"
✗ NOT "I completed my checklist"
✗ NOT "I tried my best"
✓ Production-ready. Deploy-and-forget quality.
✓ Real-world working. If it's marketing, it converts.
✓ If it's code, it's in CI/CD, tested, documented, complete.
✓ A human could ship this TODAY and never touch it again.

Documentation

• How to Use Autopoiesis MCP - Complete usage guide for standalone and PAIA integration
• How I Integrated Ralph into PAIA - Technical deep-dive on the architecture

PAIA Integration

Autopoiesis integrates with the PAIA (Personal AI Agent) compound intelligence system:

• STARSHIP - Course plotting and flight configs
• WAYPOINT - Step-by-step flight execution with automatic autopoiesis reminders
• STARLOG - Session tracking with debug diary context
• OMNISANC - Mode-aware prompt injection

When integrated, the stop hook reads system state and provides contextually appropriate prompts based on your current course, waypoint step, and recent work.

See the usage guide for details.

Environment Variables

Variable	Default	Description
AUTOPOIESIS_ACTIVE_PROMISE_PATH	/tmp/active_promise.md	Active promise file location
AUTOPOIESIS_BLOCK_REPORT_PATH	/tmp/block_report.json	Block report file location
AUTOPOIESIS_TMP_DIR	/tmp	Template vendoring directory
HEAVEN_DATA_DIR	/tmp/heaven_data	PAIA data directory (for integration)

Philosophy

The name comes from the biological concept of autopoiesis - self-maintaining systems that produce and maintain themselves.

A living cell maintains itself through metabolic processes. This system maintains itself through honest work. The loop continues until:

1. Genuine completion - Work meets production standards
2. Genuine blockage - External help is truly needed

There is no third option. Lying about completion kills the system's integrity. Lying about blockage kills trust. The only survival is honesty.

Credits

• Ralph Wiggum technique by Geoffrey Huntley
• Anthropic's Ralph Wiggum plugin for Claude Code
• PAIA compound intelligence architecture

License

MIT