CCortex AgentiX

Cortex AgentiX is Palo Alto Networks’ platform for running AI agents in security operations and IT. It replaces Cortex XSOAR as the company’s automation product, ships prebuilt agents for threat intelligence, email, endpoint, network, cloud and IT work, and lets teams build their own. Palo Alto Networks says the agents were trained on 1.2 billion real-world playbook executions and claims up to a 98% cut in mean time to respond. It is best for large SOCs, particularly existing Cortex XSIAM, Cortex Cloud and XSOAR customers.

Pricing and setup effort are the main points to plan for. No prices are published, it launched embedded in larger Cortex platforms, and getting value means designing which actions agents may take on their own. In return, buyers get more than 1,000 integrations and governance strong enough for FedRAMP Moderate and High authorization. This review is based on Palo Alto Networks’ announcements, blog posts and documentation as of October 2026; we have not tested it hands-on.

Published Scale Figures Behind AgentiX

Palo Alto Networks has published the scale behind AgentiX but not its price or customer count. Each item below was checked on the linked page.

The 1.2 billion playbook executions are the most distinctive number. Palo Alto Networks has sold security automation for more than a decade, and that history is what the agents learned from. If it holds up, AgentiX agents should handle common incident types with fewer surprises than agents trained on synthetic data. The 98% and 75% figures are best-case vendor claims, so ask for references in your sector.

FedRAMP Moderate and High authorization matters to public sector and regulated buyers, who can now adopt agentic automation without building their own approval case. Model choice per task also matters: teams can trade cost and speed against reasoning quality.

Agents on Top of Playbooks, Scripts and Packs

AgentiX combines AI agents with the automation machinery XSOAR customers already know: integrations, playbooks, scripts and Marketplace content packs. Agents can use existing playbooks, scripts and commands as actions, and playbooks can include an AI prompt step that calls a model and returns structured results.

Prebuilt Agents

• Threat Intelligence Agent: gathers and enriches threat intelligence to find related cases and emerging attacker techniques.
• Email Investigation Agent: handles email threats from search and analysis through containment across platforms.
• Endpoint Investigation Agent: runs analysis, collects forensics and contains hosts across major EDR platforms.
• Network Security Agent: manages threat response and policy across Palo Alto Networks and third-party firewalls.
• Cloud Security Agent: covers cloud posture, application protection and detection and response.
• IT Agent: automates upgrades, patching, troubleshooting and user onboarding.

An Automation Engineer agent also writes Python automation scripts from natural-language prompts in the Agentic Assistant.

A Typical Investigation

1. Goal. An analyst prompts an agent in plain language, or a trigger starts it.
2. Plan. The agent reasons about the incident and builds a plan, which Palo Alto Networks says matters most for incidents with no predefined playbook.
3. Act. It gathers data and takes actions through integrations, within the same permissions a human analyst would have.
4. Approve. Actions marked sensitive, such as changes to production systems, wait for a quick human confirmation.
5. Record. The reasoning, plan and actions are logged and auditable, and the Command Center shows agents, executed plans, triggers, prompts and open cases.

Building Custom Agents

Teams build their own agents with a GenAI builder, reusing existing automation as actions. Native MCP support lets agents connect to tools that speak the Model Context Protocol. As of the July 2026 release, teams choose Claude Sonnet 4.6, Claude Opus 4.8 or Gemini 3.5 Flash per task, and Extended Threat Intel combines Unit 42 intelligence with organizational context.

Handling Incidents Without a Playbook

Classic SOAR breaks down when an incident does not match any playbook someone wrote in advance. Palo Alto Networks’ Gonen Fink made this point at launch: if you have never seen an incident before, you do not have a playbook for it. AgentiX adds reasoning, context and planning for those cases. The agent works out the steps, uses whatever integrations and existing actions fit, and shows its plan so analysts can follow along. For the large share of routine incidents that do have playbooks, those playbooks keep running, and agents can call them as tools.

One Prebuilt Agent First, IT Ops Second

Palo Alto Networks positions AgentiX first at the SOC, with IT operations as the second target. A sensible rollout starts with one prebuilt agent on a well-understood workload, such as phishing email investigation, where actions are easy to review. Teams then mark containment and production changes as sensitive, watch the Command Center for plans and outcomes, and expand to endpoint, network and cloud agents once the approval rules feel right. Marketplace content packs supply ready-made AI actions, integrations, playbooks and dashboards for common use cases, which shortens this work.

The Automation Team Becomes Agent Owners

People decide which actions count as sensitive, approve those actions, design custom agents and maintain integrations. In most deployments the automation team that ran XSOAR becomes the team that governs AgentiX agents.

Packaging Per the Launch, Prices Unpublished

Palo Alto Networks does not publish AgentiX prices. The table shows how it is packaged according to the company’s launch announcement.

What an Existing XSIAM Contract Already Covers

An existing XSIAM customer. AgentiX arrived inside XSIAM, so the first question is what your current contract already covers. Ask your account team which agents and how much model usage are included before paying for anything new.

An XSOAR customer. AgentiX is positioned as XSOAR’s next generation, and it reuses playbooks, scripts and integrations as agent actions. Ask how your XSOAR licensing maps to AgentiX and what migration help is included.

A new buyer outside Cortex. The standalone platform is the route in. Get a written quote, ask how pricing scales with users, agents or actions, and compare it with independent AI SOC tools that publish clearer terms.

Hidden and variable costs. The larger cost is people: automation engineers to design agents, set approval rules and maintain integrations. Factor in professional services if your team has not run XSOAR before.

Comparing quotes. When you line AgentiX up against an independent AI SOC agent, compare like for like: the number of agents and integrations you need, whether model usage is included, the support tier, and the staff time each option demands. A turnkey analyst may cost less to run even if its license is similar, while AgentiX may replace several separate automation and orchestration tools.

Cortex AgentiX Compared With Torq Socrates, Microsoft Security Copilot and Purple AI

AgentiX competes on two fronts: with automation platforms that added AI agents, such as Torq, and with platform-native AI analysts from the other large security vendors.

Torq Socrates is the closest match: both pair an AI analyst with a broad automation platform. Microsoft Security Copilot and Purple AI are better fits for companies standardized on Microsoft or SentinelOne. AgentiX’s edge is breadth of integrations and governance, which suits large, mixed environments. Independent AI SOC agents such as 7AI and Dropzone AI are lighter options. Browse more on our cybersecurity agents page.

XSIAM, Cortex Cloud and XSOAR Teams

• Cortex XSIAM and Cortex Cloud customers: AgentiX is already part of those platforms, so start there.
• XSOAR teams: existing playbooks, scripts and integrations become agent actions.
• Public sector and regulated organizations: FedRAMP Moderate and High authorization eases adoption; see our enterprise agents page.
• Teams spanning SOC and IT: the IT Agent handles patching, upgrades and onboarding alongside security work; see our IT and DevOps agents page.
• Organizations wanting to build their own agents: the GenAI builder and MCP support suit teams with automation skills; see our agent builders page.

Small Teams Without Automation Staff

• Small teams without automation staff: a turnkey AI analyst will usually deliver value sooner.
• Buyers who need published prices: everything is quoted.
• Companies committed to another security platform: Microsoft or SentinelOne shops may get more from their vendor’s built-in agent.
• Teams that only want alert triage: AgentiX’s breadth may be more than you need.

Sensitive-Action Rules and Vendor Claims

• Governance design: deciding which actions need approval is your job; mark production-changing actions as sensitive from day one.
• Vendor claims: the 98% MTTR reduction and 75% less manual work are Palo Alto Networks figures.
• Packaging changes: standalone and XDR versions were announced for early 2026, so confirm exactly what your edition includes.
• Platform pull: the most value comes with other Cortex products, which can deepen dependence on one vendor.
• Model choice and cost: picking larger models per task may affect performance and cost; ask how model usage is billed.

Per-Task Model Choice in AgentiX 1.4

On July 30, 2026, Palo Alto Networks shipped AgentiX 1.4 with per-task model choice among Claude Sonnet 4.6, Claude Opus 4.8 and Gemini 3.5 Flash, Extended Threat Intel built on Unit 42 data, and FedRAMP Moderate and High authorization.

On October 28, 2025, the company introduced Cortex AgentiX as the next generation of Cortex XSOAR. CRN reported that it launched inside Cortex Cloud and Cortex XSIAM, with a standalone version planned for early 2026, alongside Cortex Cloud 2.0 and Prisma AIRS 2.0.

For Palo Alto Shops With Big Security Teams

For large security teams already invested in Palo Alto Networks, yes. AgentiX scores 8.1/10 on our scale. It offers one of the largest integration catalogs of any security agent platform, governance built for regulated buyers and a decade of XSOAR automation behind its agents.

For other teams the decision takes more weighing. Prices are not published, setup is a project rather than a switch, and AgentiX has no public user ratings of its own yet. Teams without automation engineers will likely get faster results from a turnkey AI SOC analyst. If you do evaluate it, start with one prebuilt agent, such as email investigation, keep sensitive actions behind approvals and measure response times before widening its reach.

How AgentiX Relates to Cortex XSOAR

Is Cortex AgentiX the Same as Cortex XSOAR?

AgentiX is the next generation of Cortex XSOAR. It keeps XSOAR’s integrations, playbooks and scripts and adds AI agents that plan and act, using existing automation as their actions.

How Much Does Cortex AgentiX Cost?

Palo Alto Networks does not publish prices. AgentiX launched embedded in Cortex XSIAM and Cortex Cloud, with standalone and Cortex XDR versions announced for early 2026. Ask sales or a partner for a quote.

Which AI Models Does Cortex AgentiX Use?

As of AgentiX 1.4 in July 2026, teams can choose Claude Sonnet 4.6, Claude Opus 4.8 or Gemini 3.5 Flash for each task. Palo Alto Networks says it will add newer frontier models over time.

Can AgentiX Agents Change Production Systems on Their Own?

Only if you allow it. Agents work within the same role-based permissions as human analysts, and actions that make real-world changes in production can be marked sensitive so they need a manual review first. Every step is logged.

Is Cortex AgentiX FedRAMP Authorized?

Yes. Palo Alto Networks said in July 2026 that Cortex AgentiX achieved FedRAMP Moderate and High authorization, which opens it to public sector and highly regulated organizations.