Virtana has added a way for its AI agents to bring agents from other vendors into IT incident investigations. The aim is to let software across a company help diagnose an outage together, rather than leaving engineers to stitch the findings together by hand.
Watch Desk analysis
What happened
Virtana’s new Agentic Observability Enterprise Management feature lets a Virtana agent invite other agents that support the Model Context Protocol or Agent2Agent protocol into an investigation. The company’s existing monitoring feed covers more than 25,000 real-time metrics, and its System Dependency Graph gives participating agents a shared view of how applications relate to the infrastructure underneath them.
In the example described by SiliconANGLE, an outside agent spots a slowdown and asks Virtana’s root-cause agent to investigate. It identifies a database bottleneck following a software deployment; other agents can then be brought in to help with a fix, and Virtana’s software checks whether the application has recovered.
Why it matters
Most enterprise agents have been built into individual vendors’ products. When an incident crosses several systems, those agents may each see only part of the problem, leaving people to assemble the picture. Virtana’s approach is to let agents from different providers collaborate around shared monitoring and dependency information.
Customers can define which kinds of incidents qualify for autonomous action. They can instead have engineers carry out recommendations, or require sign-off before each agent action. Policies also control which outside agents may join and how many tokens they can use, while an audit log records agent actions. Those are meaningful controls around a system designed to let software invite other software into an investigation.
Our read
This is a concrete attempt to make enterprise agents work across vendor boundaries, with practical details on permissions and oversight rather than just a grand promise of autonomy. The example is a product illustration, not evidence that the system has solved incident response in live customer environments. The real test is whether the shared context leads to faster, reliable diagnoses without making the control room harder to understand.
What to watch
- Which outside agents and protocols customers can use in practice.
- Whether Virtana publishes customer results on diagnosis time and successful fixes.
- How customers set the boundaries for autonomous action and review the audit trail.
Discussion spark: When an IT incident spans several vendors, should agents be allowed to investigate and act together within customer-set rules, or should cross-vendor actions always require human approval?
Sources and evidence
- Virtana lets its AI agents pull in agents from other vendors to work IT incidents – SiliconANGLE (28 October 2026, 13:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.