Information technology observability company Virtana today added Agentic Observability Enterprise Management to its platform so its artificial intelligence agents can pull agents from other vendors into an IT incident.

The release goes after a gap left by the way enterprises have adopted AI agents so far. Most enterprise agents have come built into a single vendor’s product, such as an IT service management system. When a failure spreads across systems from several vendors, none of those agents sees the whole outage. Collating the findings still falls to engineers.

With the new capabilities, a Virtana agent that spots trouble can invite other vendors’ agents into the investigation. Any agent that speaks the Model Context Protocol or the open Agent2Agent protocol is eligible, including those bundled with software a company already owns.

Outside agents have been able to query Virtana’s monitoring data since February, when the company released its System-Aware MCP Server. The feed runs to more than 25,000 real-time metrics. Agents called into an incident also get a shared view of how applications depend on the infrastructure beneath them from the System Dependency Graph, which Virtana introduced at the same time. A month later Virtana extended its monitoring to AI workloads.

In Virtana’s illustration, a third-party agent notices an application has slowed down and asks Virtana’s root cause analysis agent, which it locates through Agent2Agent, to investigate. The diagnosis points to a database bottleneck that appeared after a recent software deployment. Follow-up questions can go back to the root cause agent before other specialized agents are lined up for the fix. Virtana’s software then confirms the application is back to normal.

A conventional IT war room takes hours to staff and run, according to Chief Product Officer Amitkumar Rathi, who said a team of agents can get through the same work “in milliseconds.”

That kind of autonomy applies only to incident types a customer has defined in advance. Customers wanting more caution can have engineers carry out whatever fixes the agents recommend, or require sign-off on each agent action before it runs. Policy also decides which outside agents get into an investigation and how many tokens they are allowed to consume once there. Every agent action is written to an audit log.

Virtana, which went by Virtual Instruments until a 2019 rebrand, is a venture capital-backed company that started out selling tools for monitoring the performance of IT infrastructure. Investors put $73 million into the business in 2022, with Atalaya Capital Management LP and Benhamou Global Ventures among them.

Image: Virtana