Creativity · Agent Protocol
Manus Agent Platform
Manus, launched by the Chinese startup Monica in early 2025, is a general-purpose agent platform that runs long-horizon autonomous workflows — booking travel, drafting reports, building websites — using a combination of LLM reasoning, a controlled browser, and a sandboxed compute environment. Its launch drew heavy attention for demonstrating consumer-facing autonomous agents outside the coding niche.
Protocol facts
- Sponsor
- Monica (Manus AI)
- Status
- stable
- Spec
- https://manus.im/
- Interop with
- browser automation, MCP, A2A, sandbox compute
Frequently asked questions
What can Manus actually do?
Publicly demonstrated workflows include travel planning, resume shortlisting, website building, data analysis, and report writing. The agent autonomously decides which tools and sub-steps to run.
Is Manus a new foundation model?
No — Manus is an agent layer on top of existing foundation models (with disclosed use of models from multiple vendors). The interesting engineering is the orchestration, memory, and tool layer, not a new base model.
How does Manus fit into an A2A world?
A2A-ready deployments could expose Manus as a remote agent addressable by its Agent Card — letting other agents delegate consumer-style 'do this for me' tasks to it over a standard protocol.
Sources
- Manus — product site — accessed 2026-04-20