Add and configure AI agents in Pendo

Last updated:

This article explains how to add and configure an AI agent in Pendo. For an overview of how data capture works and how the two capture methods compare, see Overview of Agent Analytics. For information on exploring conversation data and analyzing usage trends after setup, see Analyze interactions with AI agents.

Prerequisites

To add and configure AI agents, you must:

  • Be on agent version 2.300.0 or above.
  • Be a subscription admin or assigned the AI Agent Admin role.

Important: Agent Analytics can only collect data from AI agents that run within applications where Pendo is installed and can execute code. If your AI agent is provided by a third-party platform (such as Zendesk or Intercom), Agent Analytics can only collect customer interaction data if that provider allows Pendo to be installed in the agent experience.

Add an AI agent

Start by filling out the agent details form. After you select an application, your next steps depend on how that app is configured in Pendo:

  • Web apps and mobile apps route you directly to Conversations API setup.
  • Browser extension apps require you to choose a capture type (prompts only or full conversations) before continuing.

To add an agent:

  1. Go to Product > Agent Analytics, then select + Add an agent.
  2. Fill out the agent details form. Together, these fields give Pendo the context it needs to tailor AI-powered insights and analysis for this agent. Clear, specific details improve how prompts are grouped and how insights are surfaced in Agent Analytics. You can update all of these fields later from the agent's settings.
    • Agent name. A label that helps you identify this agent across Pendo.
    • Agent role. A short description of what the agent does.
    • Primary users. Who interacts with this agent.
    • Agent purpose. Describe the job the agent does for users in as much detail as possible — for example, what questions it answers, what tasks it helps with, and how it does so. The more specific you are, the better Pendo can group related prompts and surface relevant insights.
    • Application. The application where this agent lives. This selection determines your next steps.
  3. Select Add agent, then follow the steps for your app type above.

Set up the Conversations API

This section applies to web and mobile apps, and to browser extension apps set up to capture full conversations.

Full conversation capture requires a developer to implement the Conversations API before data collection can begin. No data will appear in Pendo until the implementation is complete.

  1. In the Add an agent wizard, select Copy setup instructions. This copies a ready-to-use ticket description that includes:
    • The Agent ID (pre-populated).
    • A link to the full setup instructions.
    • Two implementation options: automating setup with the Pendo setup-agent-analytics AI skill (for Claude Code or Cursor users), or sending events from your backend using the server-side Conversations API.
    • Acceptance criteria for the developer.

      AgentAnalytics_AddAgent_CopySetupInstructions.png
       
  2. Paste the copied description into a ticket in Jira, Linear, or wherever your team tracks development work, and assign it to your engineering team.
  3. Select Close to exit the setup flow.

Important: Your agent is configured in Pendo, but no data will appear until a developer completes the Conversations API implementation. We recommend sending the setup instructions to your developer as soon as possible.

After your developer deploys the changes, go to the Conversations tab in Agent Analytics to confirm data is appearing, or see Test your agent setup. New conversations typically show up within one hour. For full implementation details, see Capture full conversations with the Conversations API.

Tag your agent

This section applies to browser extension apps if you selected Prompts only as the capture type.

Tagging an agent works similarly to tagging Pages and Features with the Visual Design Studio, but agents are treated as a separate event type and data isn't retroactive.

Step 1. Launch the Visual Design Studio

  1. Enter the URL where your AI agent is located in your app.
  2. Choose Tag side-by-side or Tag in an overlay, then select Launch Designer to open the designer.

Step 2. Tag the prompt input field

  1. Under Input field, choose Select.

    AgentAnalytics_TagAgent_SelectInput.png
     
  2. Select the prompt input field in your app.
  3. After you tag the prompt input field, use the Rule builder or Custom CSS options to refine the rule as needed.

Step 3. Tag the submission action

If your agent doesn't have a submit button, skip to the next section, Step 4. Define tag settings.

  1. In the Submit button section, choose Select.

    AgentAnalytics_TagAgent_SubmitButton.png
     
  2. Select the prompt submission button in your app. This captures both keyboard submissions (for example, pressing Enter or Return) and manual selections of the button.
  3. After you tag the submit button, use the Rule builder or Custom CSS options to refine the rule as needed.

Step 4. Define tag settings

  1. Under Settings, select whether the agent persists Sitewide in your application or Only on this page.
  2. Optionally, open and set the Advanced Settings:
    • Filter dynamic custom HTML attribute values. Filters out dynamic-looking values for the custom attributes your subscription admin defined under Collect custom HTML attributes in your application settings.
    • Filter other dynamic elements on selection. Filters out dynamic-looking values for standard HTML attributes Pendo collects (for example, id and class).
  3. Select Save to start capturing prompts for the agent. Prompt data isn't collected retroactively and only includes activity after tagging.

    AgentAnalytics_TagAgent_Save.png

Test your agent setup

After creating or updating an AI agent, test that Pendo is capturing data correctly.

To verify prompt capture:

  1. Wait a few minutes after creating your agent to allow the configuration to process.
  2. Open a new browser tab and open the page where your AI agent appears. This ensures you're loading the latest Pendo install script.
  3. Right-click on the page, then select Inspect, and choose the Console tab to open your browser's developer console.
  4. Open the Pendo debugger by running pendo.enableDebugging().
  5. Open the Events tab, then select Start event logging.

    Debugger_StartEventLogging.png
     
  6. Submit a prompt to your agent and verify that you see it show up as a prompt event in the logs.

    Debugger_PromptEvent.png

If no prompt event appears:

  • Wait 15 minutes and test again to make sure your new agent's configuration has been applied to the Pendo install script.
  • If it's still not showing up, double-check your installation, as there may be a problem with your setup.
  • If the problem persists, contact our support team.

View and manage agent settings

To review or update an agent:

  1. Go to Product > Agent Analytics.
  2. Select the settings (cog) icon next to the agent name.

    AgentAnalytics_AgentSettings.png

    Alternatively, select the agent from the list, then select Actions > Edit settings from the top-right corner of the agent's details page.

    AgentAnalytics_AgentDetails_EditSettings.png

The settings page includes the:

  • Agent name.
  • Agent role.
  • Primary users.
  • Agent purpose.
  • Application the agent is assigned to.
  • App type (web app, mobile app, or browser extension).
  • Tagging rules and option to Open designer (for prompts-only agents).
  • Conversations setup section, including:
    • Agent ID.
    • Install snippet example.
    • Option to Activate full conversations (for prompts-only agents) or Pause conversation collection (for agents set up to capture full conversations).
    • View installation instructions link.
  • User who added the agent and when.
  • User who last updated the agent and when.
  • Option to delete the agent.

Note: If an agent set up to capture full conversations hasn't received any data for 14 days, a banner appears in agent settings with the option to switch back to prompts-only mode.

AgentAnalytics_AgentSettings.png

Switch a prompts-only agent to full conversations

If you've already created a prompts-only agent and want to capture full conversations, you can activate conversation mode from the agent's settings.

  1. Go to your agent's settings.
  2. For Conversations, change the dropdown selection to Active.

    AgentAnalytics_Settings_Conversations.png

Important: Switching to full conversations is permanent. After activation, you can't return to prompts-only capture.

After activating conversations, your developer must implement the Conversations API to send prompts and responses to Pendo. Data won't appear until the required code changes are complete.

Assign the AI Agent Admin role

By default, subscription admins can access Agent Analytics and view the specific prompts that visitors submit to your conversational UI after it's tagged as an AI agent in Pendo.

If you're a subscription admin and want a non-admin user to view prompt text, you can assign them the AI Agent Admin role. This role allows the user to:

  • View visitor-submitted prompts and conversations.
  • View and edit settings for tagged AI agents.
  • Tag new AI agents and delete existing ones.

To assign the role:

  1. Go to Settings > Users and teams from the left-side menu.
  2. Find the user you want to assign this role to. Hover over their row in the Users table, then select the edit (pencil) icon at the far right.

    Users_Edit.png
     
  3. In the APP-LEVEL ROLES section, make sure each application to which you want this role assigned is checked in the left panel.
  4. In the right panel, select AI Agent Admin.
  5. Select Save changes, then repeat these steps for each user you want to assign this role to.
Was this article helpful?
3 out of 4 found this helpful