Palo Alto Networks
Also known as: PANW, Palo Alto, Cortex AgentiX, AgentiX, Cortex XSIAM, Cortex XSOAR, Cortex Cloud, Cortex XDR, Prisma AIRS, Unit 42, Cortex Extended Data Lake
Cortex AgentiX, the successor to XSOAR, runs prebuilt and custom security agents over 1,100 integrations and 1,300 playbooks, with manual approval for sensitive actions, every agent action logged, a model selector (Gemini, Claude) and tenants in ten regions.
Palo Alto Networks is a cybersecurity company whose agent platform is Cortex AgentiX, the successor to Cortex XSOAR. AgentiX builds, deploys and governs AI agents for security operations, cloud security and IT. It ships system agents, including Case Investigation, Automation Engineer, Phishing Response, Threat Intel and Cloud Posture agents, and lets teams build custom agents from natural-language prompts, grounded in their own context and policies. Agents draw on more than 1,100 integrations and more than 1,300 playbooks, AI prompts can be dropped into playbooks as steps, and integrations can call external MCP servers. A Cortex MCP server lets LLM clients query the platform.
Administrators decide when agents act on their own and when they need approval, and actions marked as sensitive require manual approval. Every action an agent executes is logged, along with its reasoning and plan. A model selector gives EU and US tenants a choice of Gemini 3.5 Flash, Claude Sonnet 4.6 and Claude Opus 4.8. Tenants are available in the US, the EU, Singapore, Japan, India, the UK, Canada, Germany, Australia and South Korea. AgentiX runs embedded in Cortex XSIAM, XDR and Cortex Cloud and as a standalone platform, and its agents were trained on 1.2 billion playbook executions.
Vendor details
Canonical URL
https://www.paloaltonetworks.com
Category
Security / SOC agent
Subcategory
Agentic security operations platform
Funding status
Public company (NASDAQ: PANW). Its identity business is CyberArk, acquired in February 2026 and now marketed as Idira.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
More than 1,100 prebuilt integrations and 1,300 playbooks, integrations that call external MCP servers, a Slack integration to trigger agents, and a Cortex MCP server for LLM clients. AgentiX runs natively in Cortex XSIAM, Cortex XDR and Cortex Cloud, and the Cortex API documentation includes AgentiX APIs.
In practice
A SOC replaces manual triage with prebuilt agents that plan, reason and execute a full response workflow end to end, against playbooks the organization already trusted under XSOAR.
A cloud security team lets agents investigate and resolve posture and detection issues autonomously, with guardrails ensuring every automated action stays inside the organization's existing policy.
A platform team writes a custom agent in natural language over the 1,100 existing integrations, so no new connectors need writing, and assigns tools from external MCP servers to reach systems outside the Cortex estate.
Sources & related URLs
Research sources
Agentic Index coverage score
10.5 / 14 capabilities · 75%
| Integrations & Tool Calling | Full |
|---|---|
|
More than 1,100 prebuilt integrations connect through APIs, and some integrations call external MCP servers. A Slack integration triggers agents and runs remote executions, and agents deploy natively across Cortex XSIAM, XDR and Cortex Cloud. Admins can also add MCP integrations in the agents hub as extra tools for agents, and AgentiX powers the Cortex Agentic Assistant inside each of those products. SourcePalo Alto Networks, paloaltonetworks.com/cortex/agentix and cortex-docs.paloaltonetworks.com agents hub MCP integrationsread 2026-10-05 |
|
| Workflow Orchestration | Full |
|
Several system agents do the work (Case Investigation, Automation Engineer, Phishing Response, Threat Intel and Cloud Posture), and customers can build their own from natural language prompts. They run over more than 1,300 playbooks, where AI prompts can be dropped in as steps. Agents act through four kinds of actions, playbooks, scripts, commands and AI prompts, and custom agents can run custom scripts. Admins register actions, build agents and assign actions to them in the Agentic Assistant Hub. SourcePalo Alto Networks, paloaltonetworks.com/cortex/agentix and cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiX and Configure the Cortex Agentic Assistantread 2026-10-05 |
|
| Knowledge Grounding & RAG | Partial |
|
Custom agents are grounded in organizational context and policies, and agents work over the customer's case and playbook data. Palo Alto does not describe how agents search that material. Palo Alto says its agents were trained on 1.2 billion playbook executions. In the Agentic Assistant Hub, admins can add knowledge sources that give agents business context, and a Help Center agent answers questions from the product documentation. SourcePalo Alto Networks, paloaltonetworks.com/cortex/agentix and cortex-docs.paloaltonetworks.com Configure the Cortex Agentic Assistantread 2026-10-05 |
|
| Human Oversight & Guardrails | Full |
|
Actions marked as sensitive require manual approval. Administrators decide when agents act on their own and when they need approval, and any sensitive action can be flagged to require a person's validation before a critical system change runs. Agents are also bound by the organization's existing roles and permissions. An agent never exceeds the permissions of the user running it, so approvals sit inside the access the organization already grants. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiX and Configure the Cortex Agentic Assistantread 2026-10-05 |
|
| Security, Identity & Governance | Full |
|
Attestations include SOC 2+ reports, ISO certifications and FedRAMP for cloud products, though they are not mapped to Cortex products one by one. Role based access controls bind agents to the organization's existing roles and permissions. The same access controls decide who can build agents and register actions in the Agentic Assistant Hub, and an agent cannot do more than the user running it is allowed to. SourcePalo Alto Networks, paloaltonetworks.com trust center certifications and cortex-docs.paloaltonetworks.com Configure the Cortex Agentic Assistantread 2026-10-05 |
|
| Observability & Auditability | Full |
|
All actions an agent executes are logged, and the agent's reasoning, its reading of the request and the plan it creates are visible alongside the actions it takes, with every step logged and fully auditable. The logs cover every kind of action an agent takes, whether a playbook, a script, a command or an AI prompt. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiXread 2026-10-05 |
|
| Memory & State Persistence | Not documented |
|
The platform has data stores, but Palo Alto describes no memory that agents keep from one run to the next. The Agentic Assistant is made up of agents, actions, knowledge and access control, with no memory component among them. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiX and Configure the Cortex Agentic Assistantread 2026-10-05 |
|
| Deployment & Data Residency | Full |
|
Tenants sit in named regions, the US and CA in North America, the EU, UK and DE in Europe, and SG, JP and IN in Asia. Which models are available depends on the region. Australia and South Korea are also listed, with the EU tenant in the Netherlands, and the wider model choice is offered in the EU and US regions. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiXread 2026-10-05 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Five named agents come prebuilt (Case Investigation, Automation Engineer, Phishing Response, Threat Intel and Cloud Posture), and each does its own job. More than 1,300 playbooks ship alongside them. There is also a Help Center agent for questions about the product, and the Threat Intel agent supports threat hunting. SourcePalo Alto Networks, paloaltonetworks.com/cortex/agentix and cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiXread 2026-10-05 |
|
| Triggers & Channel Coverage | Full |
|
Agents run around the clock on cases and alerts inside XSIAM, XDR and Cortex Cloud, and playbooks run agent steps as incidents arrive. Slack can trigger agents too, so alerts wake them and chat reaches them. In the Cortex Agentic Assistant, people start an agent with a plain language prompt. AI prompts placed in a playbook run as their own workflow steps, and from Slack a user can also investigate and run remote executions. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiX and paloaltonetworks.com/cortex/agentixread 2026-10-05 |
|
| Model Flexibility & Routing | Full |
|
A model selector gives EU and US tenants a choice of Gemini 3.5 Flash, Claude Sonnet 4.6 and Claude Opus 4.8, with Claude Sonnet 5 on request, while other regions have Gemini 3.5 Flash. The customer makes the choice, and each region has its own list of models. The selector labels the three tiers Flash, Thinking and Pro, and the choice is made per chat. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiXread 2026-10-05 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Cortex AgentiX APIs sit next to the API references for the other Cortex products, XSIAM, XDR, XSOAR and Cortex Cloud, with API key generation. A Cortex MCP server lets LLM clients query the platform. The MCP server is a download that runs on a local machine or in a container, and it works with Claude Desktop as the client while supporting any MCP client, for investigating and managing cases and issues and for building custom tools. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Cortex API overview and Cortex MCP serverread 2026-10-05 |
|
| Testing, Debugging & Optimization | Not documented |
|
Customers get no published way to test agents and score the results. Palo Alto reports training on 1.2 billion playbook executions and MTTR figures from its own development and results, and says more than 1,000 Cortex customers have turned AgentiX on. SourcePalo Alto Networks, paloaltonetworks.com agentic workforce blog and /cortex/agentixread 2026-10-05 |
|
| Browser & Computer Use | Not documented |
|
Agents act through integrations, playbooks and scripts, with no browser or computer use. Each agent action is a playbook, a script, a command or an AI prompt, and agents can also visualize data or an investigation. None of those actions opens a browser or works a desktop screen. SourcePalo Alto Networks, cortex-docs.paloaltonetworks.com Agentic AI in Cortex AgentiXread 2026-10-05 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Cortex Cloud now plugs into Claude Code and Cursor through the Cortex Cloud AppSec MCP Server, listed in the Claude plugin directory and the Cursor Marketplace. Coding agents pull SAST secure coding rules before generating code, check open source packages for malicious or suspicious dependencies before install and can block them, and run secrets, package and IaC scans before and after writing files.
Bears on: Human approval / guardrails
View sourcePalo Alto Networks launched Unit 42 Continuous Frontier AI Defense, an agentic security service available worldwide by annual subscription. It continuously tests changing environments, validates attack paths across applications, APIs, cloud infrastructure and other assets, and supplies prioritized remediation guidance. A routing system assigns security tasks to frontier and open weight models.
Bears on: Security / enterprise
View sourcePalo Alto Networks has acquired Console, an AI-native platform designed to deliver autonomous agentic capabilities across enterprise operations. The acquisition will integrate Console's technology into the Cortex security operations portfolio to automate security investigation, prioritization, and response workflows.
Bears on: Agent capability
View sourcePricing
Sold through sales, with no published price. AgentiX comes embedded in Cortex platform agreements or as the standalone AgentiX platform.
An enterprise platform agreement. The billing unit is not disclosed.
What is public
Product packaging and availability, embedded and standalone, are public. No prices are published.
Billing mechanics
Billing terms are not disclosed. AgentiX is sold through direct enterprise sales and the channel, structured around multi product platform agreements rather than per product licenses.
Cost watchouts
When AgentiX is bought inside a Cortex platform agreement, no separate price for it is published, so its cost sits inside the platform commitment.
Variable cost rationale
No billing unit is published, so nothing public says what is metered. Cost is set by the platform agreement and its scope, whether AgentiX comes embedded in a Cortex XSIAM, XDR or Cortex Cloud agreement or as the standalone platform, and no cap or included volume is published.
Additional watchouts
No published figure separates the cost of AgentiX from the Cortex platform commitment it is bundled into.
Sales call required
Yes, required for paid access
Free / trial
None published
Lowest paid plan
Not published
Commercial notes
Palo Alto Networks is a public company (NASDAQ: PANW).
Key ambiguities
How the standalone AgentiX platform is metered and how it is priced against the embedded version are not published.
Missing data
All pricing and the AgentiX metering unit.
Related vendors
- 7AI — Agentic SOC from the Cybereason founders: domain specialized agents…
- Abnormal AI — Behavioral AI email security with three named autonomous agents: a…
- AirMDR — AI native MDR whose AI analyst investigates every alert and writes…
- Airrived — Agentic OS that unifies SOC, GRC, IAM, vulnerability management, IT,…
- Andesite — Human-AI SOC platform where customers build their own agents and…
- Anvilogic — Agentic security operations platform that works on top of existing…
Alternatives to Palo Alto Networks
The closest documented capability profiles to Palo Alto Networks among security and SOC agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Swimlane10.5 / 14Matches Palo Alto Networks across all 14 documented capabilities
- Ghost Security10.0 / 14A lighter documented profile than Palo Alto Networks
- D3 Security (Morpheus)10.5 / 14Adds documented Memory & State Persistence
- SentinelOne10.5 / 14Fuller documented coverage on Knowledge Grounding & RAG
- Tines11.5 / 14Adds documented Memory & State Persistence and Testing, Debugging & Optimization
- Andesite9.0 / 14A lighter documented profile than Palo Alto Networks
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded