Recover long-running agent runs after interruptions

Background runs streamed over Server-Sent Events can now reconnect and resume after a disconnection or page refresh. Operators rejoin the run exactly where they left off, with full context preserved.

Long-running agents and teams are common in production, particularly for research, analysis, and multi-step automation. Until now, a transient network drop or browser refresh meant losing the run or restarting from the beginning. The new behavior eliminates that failure mode, making AgentOS more reliable for the workflows users actually run on it.

For operators monitoring live agent activity, this also means fewer interrupted sessions and less wasted compute spent regenerating progress that was already complete.

Learn more in the Background Execution docs.