Manus
AI · AI
Agent tasks, their messages, files, and scheduled runs under that person’s own account.
Connecting Manus in practice
Manus rate-limits per user rather than per key — every API key one user owns shares a single counter — and the endpoint that actually starts work is the scarcest: task.create and task.sendMessage allow 10 requests a minute against 100 for the read endpoints. Tasks run asynchronously, so the polling has to fit inside the same budget.
Rate limits
- task.create and task.sendMessage: 10 per minute.
- Write endpoints such as task.update, task.stop, task.delete, project.create, file.upload and webhook.create: 40 per minute.
- Read endpoints such as user.me, task.detail, task.list, task.listMessages and connector.list: 100 per minute.
- usage.list, usage.teamStatistic and usage.teamLog: 600 per minute.
- All API keys that belong to the same user share a single counter; overage is HTTP 429 with the error code rate_limited.
Worth knowing
- Adding keys does not buy throughput, because the counter is per user, so scaling means more users rather than more credentials.
- task.create returns before the work is done. Progress comes from polling task.listMessages and follow-ups from task.sendMessage, both of which draw on the 10-per-minute write budget.
- The documented advice on 429 is exponential delay plus jitter rather than an immediate retry, and no Retry-After is documented to key off.
- Authentication is either an API key or an OAuth2 token, and an Open App acting on behalf of team users is a separate OAuth2 path from a personal key.
Checked against Manus API rate limits (2026-09-29), Manus API introduction (2026-09-29), Manus API overview (2026-09-29)
What an agent can do
Each action is granted on its own. An agent allowed to read is not thereby allowed to write, and the scope beside each row is what the acting user must have connected for it to run at all.
Often connected alongside
Put Manus behind one governed endpoint.
Same permissions, same audit trail, whatever else you connect next.