> 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-cloud-posture-management/cases-and-issues/analyze-and-resolve-cases/additional-case-actions/dispatch-an-agent-on-a-case.md).

# Dispatch an agent on a case

Dispatch AI agents to one or more cases to investigate, monitor, and act on your behalf, either once or on a recurring schedule.

Agents in Cortex Cloud can work as members of your case team. When you dispatch an agent to a case, you give it a prompt that describes its task, and the agent runs that task in the context of the case. You can run the task once, or set it to recur at a fixed interval, for example to check a case for new issues every 30 minutes.

Each dispatched agent joins the case team and is listed with the other team members. You can dispatch different agents to the same case, and you can dispatch the same agent to many cases at once. Each case gets its own task. You can follow each agent task in the case's **Resolution Center**.

### Requirements

The **Dispatch Agent** action is available only when all of the following are true:

* **Agents & LLM Experience is enabled** for the tenant. An administrator turns it on under **Configurations > Server Settings > AI Configuration**. This setting also enables the Agentic Assistant.
* **The agent is public and enabled.** Only public agents with an enabled status appear in the agent list. Private agents can't be dispatched to cases.

{% hint style="info" %}
An agent works with the permissions of the user who dispatched it, so it can see and act only on data that you can access.
{% endhint %}

### Track agent tasks in the Resolution Center

Agent tasks appear in the case's **Resolution Center** on the same tabs as playbook tasks:

<table><thead><tr><th width="124">Tab</th><th>What you see</th></tr></thead><tbody><tr><td><strong>Pending</strong></td><td>Agent tasks that are waiting for your input, such as approval for a sensitive action.</td></tr><tr><td><strong>In Progress</strong></td><td>Agent tasks that are currently running, including recurring tasks that have upcoming runs.</td></tr><tr><td><strong>Done</strong></td><td>Agent tasks that have finished.</td></tr></tbody></table>

Click an agent task to open its session. The session is the conversation thread for the task. It shows the agent's reasoning, the actions it took, and its results. Any case team member can open the session and continue the conversation with the agent.

#### Tasks with sensitive actions

Before an agent performs a sensitive action, it stops and asks for approval. The task moves to the **Pending** tab and waits there until a user approves or cancels the action. These tasks don't time out, so review the context before you approve an action that has been waiting for a long time.

#### Recurring tasks

A recurring task runs at the interval you set, for the number of runs you set. If a previous run is still in progress or waiting for input when the next run is due, Cortex Cloud skips that run.

You can cancel a recurring task from the **Resolution Center** at any time. Canceling stops all future runs. Runs that have already completed stay on the **Done** tab.

### How to dispatch an agent

1. **Open the Dispatch Agent action.**

   1. Go to the **Cases** page. View the page in the default **Table** view. You can switch views from the **Display** menu.
   2. Select one or more cases to dispatch the agent to.
   3. Right-click the selection and select **Dispatch Agent**. You can also click the **Dispatch Agent** button in the table toolbar if multiple cases are selected.

   If **Dispatch Agent** is unavailable, hover over it to see the reason:

   | Tooltip                                         | Reason                                                                                                   | What to do                                                                                                                                                    |
   | ----------------------------------------------- | -------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------- |
   | You cannot perform this action on blocked cases | At least one of the selected cases is partially permitted: it includes data outside your assigned scope. | <p>Clear those cases from your selection.</p><p>To hide them from the table, select <strong>Hide partially permitted cases</strong> in the table toolbar.</p> |
   | Dispatch Agent is not enabled for this tenant   | Agentic capabilities are not enabled for the tenant.                                                     | Ask an administrator to enable AI Configuration on the tenant in server settings.                                                                             |
   | Agents & LLM Experience is disabled             | Agents aren't enabled for the tenant.                                                                    | Ask an administrator to enable **Agents & LLM Experience** in the **AI Configuration** server settings.                                                       |
   | Dispatch Agent is unavailable for child tenant  | You're viewing a child tenant's case from a parent (MSSP) tenant.                                        | <p>Dispatch the agent from within the child tenant.</p><p>For more information, see <a href="#limitations-for-mssp-tenants">Limitations</a>.</p>              |

   If more than one reason applies, the tooltip shows the first one in the order listed above.
2. **Select an agent.**

   In the **Select agent** window, search for or browse the available agents, select one, and click **Next**.
3. **Configure the agent's task.**

   In the **Configure&#x20;*****agent name*****&#x20;agent prompt** window, enter the following:

   <table><thead><tr><th width="138">Field</th><th width="103.9998779296875">Required</th><th>Description</th></tr></thead><tbody><tr><td><strong>Name</strong></td><td>Yes</td><td>A name for the task. This name identifies the task in the <strong>Resolution Center</strong>.</td></tr><tr><td><strong>Prompt library</strong></td><td>No</td><td>Select a saved prompt to start from. The prompt text fills the <strong>Prompt</strong> field. If you then edit the prompt, the selection clears. This field appears only if prompts are enabled for your tenant.</td></tr><tr><td><strong>Prompt</strong></td><td>Yes</td><td>Describe the task that you want the agent to perform, in natural language. For example: <em>Summarize any new issues added to this case and flag anything critical.</em> Click <strong>Prompt helper</strong> for help with writing your prompt. If your tenant has model selection enabled, you can also choose a model here.</td></tr><tr><td><strong>Inputs</strong></td><td>Yes, if the prompt contains any</td><td>Type <code>$</code> in the prompt to add an input, which is a named value that the prompt uses. Each input appears under <strong>Inputs</strong>. Click an input to set its value. You must give every input a value before you can dispatch the agent. Case dispatch doesn't support issue variables, so enter static values.</td></tr><tr><td><strong>Recurring</strong></td><td>No</td><td>Turn on to run the task more than once. When this option is off, the task runs once.</td></tr><tr><td><strong>Occurrences</strong></td><td>Yes, if <strong>Recurring</strong> is on</td><td>How often the task runs: <strong>Every 10 Minutes</strong> (default), <strong>Every 30 Minutes</strong>, <strong>Every Hour</strong>, or <strong>Every 2 Hours</strong>.</td></tr><tr><td><strong>Quantity</strong></td><td>Yes, if <strong>Recurring</strong> is on</td><td>The total number of runs, from 2 to 12. The default is 2.</td></tr></tbody></table>
4. **Dispatch the agent and track its task.**
   1. Click **Dispatch**.

      Cortex Cloud creates a separate task for each selected case and shows a notification with the result. If some of the tasks fail, the notification shows how many were dispatched successfully, for example *2 of 3 agents dispatched successfully. Some failed.*
   2. Open the case and go to the **Resolution Center** to follow the agent's progress. The agent also appears in the case team.

### Limitations for MSSP tenants

On MSSP parent tenants, the following limitations apply:

* **Agents as case collaborators is always turned off.** You can't dispatch agents to cases from a parent tenant.
* **Child tenant agent tasks open in the child tenant.** You can see agent tasks from child tenants, but clicking a task redirects you to the child tenant.
* **Child tenant agents are listed by ID.** Agents from child tenants are identified by their agent ID, not their display name.


---

# 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 by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://cortex-docs.paloaltonetworks.com/cortex-cloud-posture-management/cases-and-issues/analyze-and-resolve-cases/additional-case-actions/dispatch-an-agent-on-a-case.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.
