The First L3 Agentic Orchestration Platform.
Monoco is a Headless Operating System designed to shift AI development from L2 (Agents/Chatbots) to L3 (Autonomous Engineering Systems). It provides the governance, state management, and environment isolation required to turn raw LLM capabilities into a reliable, verifiable, and autonomous workforce.
While traditional AI tools (L2) focus on turn-based chat and "helpful assistants," Monoco orchestrates Autonomous Sessions.
| Feature | L2: AI Agents (e.g. Cursor, Claude Code) | L3: Autonomous Systems (Monoco) | Value |
|---|---|---|---|
| Interaction | HITL: Human-In-The-Loop (Step-by-step) | HOTL: Human-On-The-Loop (Batch/Async) | 10x Productivity |
| Quality | Subjective satisfaction (Chat-based) | Objective DoD (Tests, Lints, Invariants) | Reliability |
| State | Ephemeral context (Chat History) | Persistent Identity (Issue Tickets) | Traceability |
| Environment | Shared / Volatile | Isolated / Sandboxed (Worktrees) | Non-Interference |
Monoco acts as a Distribution, bridging the gap between raw intelligence and industrial-grade engineering.
- LLM Kernel: The raw reasoning engine (Kimi, Gemini, Claude).
- Monoco Distro: The orchestration layer providing the Init System (Issue Lifecycle), Package Manager (Skills), and Security Policy (Guardrails).
- Client/DE: Your IDE (VSCode, Zed) interacting via standard protocols (LSP, ACP).
In Monoco, a task is not "finished" when the AI says so. It is finished when the System Invariants are met: all tests pass, the linter is silent, and the implementation matches the Issue's acceptance criteria.
Monoco treats Issues as Units of Work. Just as systemd manages system units, Monoco manages the lifecycle of an engineering task—from open to close, ensuring no "freelancing" and absolute traceability.
Monoco creates dedicated, isolated environments (Branches/Worktrees) for every task. This ensures the AI never pollutes your local state or clashes with human developers.
Policies are not just prompts; they are governed by code. Monoco enforces Git hooks, CI/CD gates, and automated audits to ensure the workforce adheres to your project's technical standards.
pip install monoco-toolkitTransform any repository into a Monoco-managed autonomous engineering environment.
monoco initInject your project's governance and standards into the Agent's constitution.
monoco syncStart the daemon to monitor the collaboration bus and schedule agent tasks.
monoco session start- State Engine: Markdown/YAML-based persistence (No heavy DB required).
- Communication: Event-driven architecture via local file system events.
- Protocols: Native support for LSP (Language Server) and ACP (Agent Client Protocol).
MIT © IndenScale