Start and follow agent work from any interface (Beta)
The Agents page now offers New run: choose a saved agent, describe the task, and start. Agent setup separates reusable instructions from individual tasks, with optional team settings collapsed.
The Team view explains what is happening and what to do next. Assignments lead each task card; usage and IDs are available in details. Search, state filters and pause/resume updates make larger teams easier to follow. The answer appears before technical turn history on Overview.
Developers can read the same progress, saved results and verification with nodus agentrun team NAME, its
-o json output, the API’s /team subresource, or the read-only MCP team tool. Commands and request examples
are available in the console. Hosted inference remains unavailable.
See Agents.