Vectra AI launched Vectra AI Pro on August 5, 2026, an attack-signal intelligence offering for AI agents and human analysts in security operations centers. According to the company's announcement and SiliconANGLE, the product correlates network, identity, cloud, SaaS, EDR and SASE telemetry, and exposes data through REST APIs, a Model Context Protocol server and prepackaged AI skills.
Vectra AI launched Vectra AI Pro on August 5, 2026, positioning the offering as a source of correlated attack-signal intelligence for AI agents and human analysts in security operations centers. The launch was announced at Black Hat USA, according to the company's PRNewswire release.
Correlating security telemetry
According to Vectra AI's announcement, Vectra AI Pro continuously combines network, identity, cloud, SaaS, endpoint detection and response (EDR), and secure access service edge (SASE) signals into a cross-domain view of attacker behavior. SiliconANGLE similarly reports that the product assembles telemetry from those domains into a single account for security agents rather than presenting isolated data sources.
The company describes the offering's behavioral intelligence as covering reconnaissance, credential abuse, privilege escalation, lateral movement, command-and-control, and data exfiltration. SiliconANGLE reports that the product also attributes activity to users, workloads or AI agents, ranks entities and attack paths, and includes exposure validation intended to verify that an attack path has been closed or an active attack stopped.
"Everyone has AI today. Vectra AI's advantage isn't that it uses AI. It's that it applies AI to understand attacker behavior," Hitesh Sheth, Vectra AI's president and CEO, said in the release. AJ Wiggan, security operations manager at Gamma, said in the same release that correlated network, identity and cloud context helps the company understand "the full attack story."
Integration options for security teams
SiliconANGLE reports three adoption paths for Vectra AI Pro:
- •REST APIs, a Model Context Protocol (MCP) server, and prepackaged AI skills for organizations building custom agents or copilots. - •Vectra-provided agents that triage and prioritize detections and create natural-language case summaries.
- •Consulting services intended to integrate the signals into existing customer tooling.
The prepackaged skills cover indicator enrichment, packet-capture retrieval, forensic investigation, threat hunting, and reporting, according to SiliconANGLE.
For security engineering teams, the technical significance is the product's emphasis on supplying agent workflows with correlated evidence and entity context. In comparable agentic SOC designs, an agent's reliability depends heavily on data normalization, identity resolution, provenance, access controls, and the ability to verify actions against source telemetry. Teams evaluating such integrations will still need to assess API permissions, MCP server governance, retention policies, and human approval boundaries for response actions. Neither source provided pricing, benchmark results, or independent measurements of investigation-time reduction. Those details would be needed to compare the offering against other security-data and agent-orchestration platforms.
Key Points #
- 1Vectra AI Pro combines security telemetry across network, identity, cloud and endpoint domains, giving agent workflows broader investigative context than isolated alerts.
- 2The product exposes REST APIs, MCP connectivity and packaged skills, making integration architecture a central consideration for custom SOC agent deployments.
- 3Across agentic SOC deployments, correlated evidence does not remove the need for governance, provenance controls and human approval of consequential response actions.
Scoring Rationale #
This is a notable security-platform release focused on a practical bottleneck for AI agents: reliable, cross-domain security context. It is relevant to SOC engineers and AI practitioners integrating agents with detection and response systems, although the sources provide no independent performance data or broad market-adoption evidence.
Sources #
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.