The company says the system can identify root causes, perform mitigations, or open and test a pull request before paging an engineer for review.
Cognition reports 100% end-to-end success for Devin Oncall on its SREGym evaluation, which requires agents to identify a root cause and mitigate the fault in the same run. It also says AngelList’s team found root causes 3.5 times more often with Devin Oncall than with its in-house investigation harness and that the agent handled 50% of incidents end-to-end. These are company-reported results tied to a specific benchmark and customer account; the announcement does not establish equivalent performance across production environments.
The launch makes production investigation and mitigation an explicit product surface for a coding agent, with engineer review and escalation still part of the described workflow. Source: https://devin.ai/blog/introducing-devin-oncall