Virtana’s AI agent war room targets cross-system IT incidents

Virtana launched Agentic Observability Enterprise Management on November 13, 2026, with general availability planned for the first quarter of 2027.

Virtana says the system coordinates AI agents from different enterprise platforms during IT incidents. It is available now to select customers.

The company says the software draws on more than 25,000 operational signals across applications, infrastructure, cloud systems and AI environments. It combines those inputs with its System Dependency Graph, a map of relationships among systems, to trace root causes and estimate business impact.

The launch addresses a widening operational gap. Virtana says companies are deploying growing numbers of agents across IT service management (ITSM), cloud, security and software-development tools. Pressure rises when incidents span several of those domains.

What does Virtana’s agentic war room actually do?

Virtana calls the coordination environment an “agentic war room.” The system recruits agents from third-party platforms that would otherwise work in isolation, then gives them a shared operational view.

Virtana uses Model Context Protocol (MCP), which the company says exposes enterprise-wide intelligence to participating agents. It also uses the open Agent2Agent (A2A) protocol, which lets agents discover one another, collaborate and act as a governed group.

The aim goes beyond isolated detection and diagnosis. Virtana says participating agents can conduct a joint investigation, execute remediation and confirm that a service has recovered.

Virtana’s example begins when a third-party agent spots a problem in an application. That agent calls Virtana’s root-cause analysis agent, which investigates with system-wide operational data.

The root-cause agent identifies a bottleneck linked to a software deployment, then works with other agents to carry out corrective action. This turns the war room from a monitoring feature into a coordinated response process.

Virtana Chief Executive Officer Paul Appleby said:

“We're bringing agents from across the enterprise together with the intelligence and coordination to tackle the most complex IT challenges at machine speed.”

Appleby also described Agentic IQ as the collective intelligence of agents that understand the entire business, reason across complex systems and solve problems no single agent could solve alone. He argued that this coordination enables more autonomous operations at scale.

Virtana says specialized agents can assemble around a shared problem, exchange findings and decide the next step in milliseconds. The announcement supplies no independent response-time test, incident sample or accuracy comparison, so that claim remains unvalidated.

The practical result is coordination across existing agents, not a single new AI model.

How much autonomy can customers grant?

Virtana built the system around staged autonomy. Customers choose among three progressively broader operating modes.

• Recommendation mode: Agents suggest fixes, and engineers implement them.
• Approval-based remediation (corrective action): Agents can remediate issues when a person approves the action.
• Autonomous resolution: Agents act independently for defined classes of incidents.

Enterprise policy controls determine which agents can participate, what they may do and when human approval is required. Virtana says those controls keep each transition inside the customer’s operating rules.

The company says every action is auditable, including records of agent activity, token consumption (model usage) and operating cost. Virtana presents those records as a way for administrators to review both behavior and expense.

Those controls answer a basic procurement question: who can act, on what, and with whose approval. They do not establish whether an agent’s diagnosis or proposed fix is correct.

Autonomy remains conditional, and policy controls do not prove the agents make sound decisions.

Why does fragmentation matter?

Research from Salesforce cited by Virtana says the average enterprise runs 12 AI agents, while half operate in silos. The announcement provides no study date or methodology, so the figure offers limited context.

In practice, a cloud agent may see infrastructure behavior, a development-platform agent may know recent code changes, and a security agent may track policy or threat issues. Without coordination, engineers must bridge those gaps themselves.

Virtana argues conventional war rooms rely heavily on engineers gathering evidence from separate tools and coordinating fixes manually. That workflow can slow incidents that cross infrastructure, code, security controls and service-management systems.

Virtana Chief Product Officer Amitkumar Rathi said:

“Agentic war rooms assemble specialized AI agents that share intelligence, resolve issues and disband in milliseconds.”

Rathi said traditional war rooms can take hours to assemble teams, diagnose problems and coordinate fixes. He also said MCP supplies enterprise-wide intelligence, while A2A lets agents discover, collaborate and act as a coordinated, governed workforce.

Virtana says the open-protocol approach lets enterprises bring in agents from platforms they already use. The company positions itself as the orchestration layer, meaning the coordination layer that provides a shared operational view.

That design avoids a forced tool replacement, but it also depends on third-party platforms exposing usable agents. Virtana names no supported agent catalog or partner list in the announcement.

Open protocols offer a path to interoperability, but the announcement does not identify which tools can join the war room.

What remains unproven before general availability?

Virtana says the product is available to select customers now. General availability is planned for the first quarter of 2027, with no firm launch date inside that window.

Virtana says customers can expand autonomy within defined policy controls, beginning with recommendations and moving toward approval-based and then autonomous remediation. The progression applies only to incident classes that the customer defines.

The announcement leaves four commercial and technical questions unanswered:

• It gives no pricing or subscription details.
• It names no select customers and provides no customer count.
• It does not identify which third-party platforms or agents support the protocol path.
• It provides no independent test of diagnosis accuracy, remediation success or recovery time.

Those omissions limit early comparisons with competing incident-response and observability (system-monitoring) products. Buyers also cannot determine from the announcement how many agents can run together, how quickly they respond in production or what a deployment costs.

Virtana’s first-quarter 2027 general-availability target is the next checkpoint for evidence from select customers. The announcement does not commit to independent validation or specify how response times, diagnosis accuracy or remediation outcomes will be measured.

Q1 2027 is the next checkpoint, not proof that autonomous incident resolution works at enterprise scale.