Autoheal

Improving operational work with engineer approval

Autoheal gives platform engineering teams a shared environment for agents that investigate incidents, remediate vulnerabilities and manage AI coding costs. It connects repositories, services, tickets and releases into a context graph, while access policies, budgets and approvals govern each agent. An Evaluator scores runs against downstream evidence such as review comments and CI failures. A Healer proposes changes to skills, memory, tool permissions or model choices as pull requests, checks them against historical runs, and requires an engineer’s approval before they go live.

Teams connect their development and production systems, choose an operational workflow and deploy within their own cloud and security controls. The shared context also improves the coding agents developers already use. Autoheal’s announcement includes customer accounts from Nomura Bank, where agents gather incident evidence under strict access controls, and AvidXchange, which reports faster root-cause investigation and plans broader use across its software lifecycle. Enterprise-specific small language models trained on private engineering data are a future development.