> For the complete documentation index, see [llms.txt](https://cortex-docs.paloaltonetworks.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://cortex-docs.paloaltonetworks.com/cortex-xdr-5.x/onboard-cortex-xdr/post-deployment-steps/configure-the-cortex-agentic-assistant/agents-hub/expand-agent-capabilities-with-mcp-integrations.md). # Expand agent capabilities with MCP integrations Cortex Agentic Assistant supports native interaction with external environments via the Model Context Protocol (MCP). Agentic Assistant agents can use tools from third-party MCP servers to retrieve data and perform tasks in external systems. For example, an agent can open a Jira issue or check GitHub to see if security scans in a workflow are being bypassed.. The Cortex Agentic Assistant connects to external MCP servers using streamable HTTP and supports both OAuth-based and Authless servers. The MCP server must be accessible via a URL. To communicate with third-party MCP servers, you install the relevant content pack from Marketplace and configure an integration instance. The integration connects to the third-party MCP server to discover available tools and automatically generate agentic actions. {% hint style="info" %} ### Note For agentic actions to be created from tools on an MCP server, the tools must have input and output descriptions on the MCP server. These descriptions are required for Cortex XDR to understand the tools' capabilities and the correct use cases. {% endhint %} ### MCP integrations To find MCP content packs, filter for **MCP** under **Types** in Marketplace. Examples of MCP content packs include **Cloudflare MCP**, **GitHub MCP**, and **Atlassian Cloud MCP**. You can also use the **Generic MCP** content pack to connect to MCP servers that do not have their own specific content pack. Each MCP integration includes instructions for providing the required parameters, such as the URL and authentication details. You can create multiple integration instances for each MCP integration. For example, you might configure one instance of the **GitHubMCP** integration to connect to a environment with read tools and another instance to connect to an environment with both read and write tools. In addition, the **GenericMCP** integration can be used to connect to multiple MCP servers, each with a separate integration instance. When you **Test** the integration instance, Cortex XDR verifies server connectivity. {% hint style="info" %} ### Note If you are using OAuth-based authentication, the **Test** button returns an error containing the command to run in the playground in order to test the connection. {% endhint %} All configured MCP integration instances can be viewed in the **Settings** → **Data Sources & Integrations** page. You can view each integration instance and verify the status of the connection, **Test** the connection, enable or disable the integration instance, and view the last discovery timestamp. ### MCP tool actions The integration instance checks hourly for new or changed tools exposed by the third-party MCP server. The same checks are also performed every time an integration instance is saved. All discovered tools are automatically registered as AI actions, with the type **MCP Tool**. Actions are created once per MCP integration instance. If you have multiple integration instances for the same MCP server, multiple actions are created for the same tools. The server name, the tool name, and the name of the integration instance are all included in the name of the action. Actions created through tool discovery are system actions. The actions cannot be edited, but you can enable and disable them and also select or clear the checkbox to mark the action as a sensitive action that requires manual approval. By default, all actions registered from MCP servers are marked as sensitive. {% hint style="info" %} ### Note * If an MCP tool is removed from the MCP server, the action will be unavailable due to missing content. If the tool is restored on the MCP server, the action is automatically reenabled. {% endhint %} ### Agents and permissions To use MCP tool actions, they must be added to custom agents in the **Agents Hub**. By default, all users with access to the custom agent can use all of the available tools. To restrict access to MCP tools, go to **Settings** → **Configurations** → **Data Collection** → **Integration Permissions**. You can restrict access for MCP integration instance commands to one or more roles. If you restrict access, only users in the permitted roles can use these actions. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://cortex-docs.paloaltonetworks.com/cortex-xdr-5.x/onboard-cortex-xdr/post-deployment-steps/configure-the-cortex-agentic-assistant/agents-hub/expand-agent-capabilities-with-mcp-integrations.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.