A high-performance Model Context Protocol (MCP) server and React/SOTA dashboard for OpenManus.
Tip
System Context: This is a Portmanteau Bridge that manages the subprocess lifecycle, persistent state, and task history for the OpenManus engine.
- FastMCP 3.2.0: Built on SOTA v1.20 for reliable asynchronous task management.
- Persistent Jobs: Agent tasks survive server restarts (
~/.openmanus-mcp/jobs.json). - Ecosystem Ready: Integrated positioning with Manus, OpenClaw, and Robofang.
- System Dashboard: Native monitoring UI for real-time subprocess management.
- Agentic Prompts: Native templates for rapid task generation.
- Sampling Support: Reasoning via host LLM sampling (SEP-1577).
OpenManus MCP is part of a broader autonomous agent fleet. Understanding the boundaries between Manus, OpenClaw, and Robofang is critical for production deployment.
- See docs/ECOSYSTEM.md for the full architectural map.
See the full ONBOARDING.md for detailed setup.
- Configure: Set
OPENMANUS_ROOTto your clone of OpenManus. - Start:
uv sync uv run openmanus-mcp - Explore: Open http://localhost:10769 to see your fleet.
In modern AI architecture, tools should be composable yet specialized. openmanus-mcp follows the Portmanteau pattern by providing a single, authoritative openmanus_bridge tool that handles:
- Orchestration: Starting and stopping agent runs.
- Telemetry: Real-time log and state ingestion.
- Persistence: Saving and restoring task history.
- Python 3.12+
- uv
- OpenManus Clone (The core engine)
- Local:
uv run pytest - Integration:
uv run pytest --integration(Runs real agent tasks) - CI: GitHub Actions with Ruff linting and build verification.
- SOTA v1.20
- FastMCP 3.2.0
- SEP-1577 (Agentic Workflow Protocol)
2026 Sandra Schi Alsergrund, Vienna. [Materialist Reductionism] [Socialist Infrastructuralism]