Autoheal raises $7.9 million for self-fixing AI agents

Autoheal's founders say they spent years building enterprise-grade AI and engineering infrastructure at Microsoft Corp., ThoughtSpot Inc. and Harness Inc., and it was at Harness that they concluded agents needed to be managed as code, overseen by continuously learning meta-agents. Until this announcement the startup operated in stealth mode, yet it already counted Normura Holdings Inc., AvidXchange Inc. and Empiric Earth Inc. among its enterprise customers. Normura Bank Chief Information Officer Sameer Jain said his production operations teams were overwhelmed by alerts and that Autoheal takes investigation timelines down from hours to minutes while running entirely within the bank's own cloud.

The funding lands as enterprises ship more software via AI code generation, a shift the company says has brought a rise in production incidents, vulnerabilities and token costs, and as teams adopt software factory models powered by dozens of specialized agents. Autoheal's answer is a shared engineering context graph maintained by two specialized agents, one rating and one repairing the rest. Investor Harpinder Singh of Innovation Endeavors frames the opportunity as larger than a single agent or workflow, describing Autoheal as an agent infrastructure layer.

What the company has described so far is a roadmap rather than a shipped product: Choudhury said Autoheal's goal now is to develop new reinforcement learning techniques to train customers' agents on their own private engineering data, potentially yielding enterprise-specific small language models, and he is convinced the architecture could expand into data and security engineering. Whether that materializes appears to hinge on how well those private-data techniques work in regulated customer environments, and on platform teams deciding that governing agents through one layer beats managing them separately.