> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.ada.cx/docs/optimization/testing/interactive-testing/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.ada.cx/_mcp/server. # Interactive Testing Interactive Testing lets you engage with your AI Agent in real time, so you can experience exactly how it responds to your end users' inquiries. > **Tip** > > Interactive Testing differs from [Simulations](/docs/optimization/testing/simulations), which run saved test cases at scale with automated pass/fail evaluation against expected outcomes. > > Use Interactive Testing for quick checks, multi-turn exploration, and qualitative validation. ## Overview Before deploying changes to end users, you can use [Interactive Testing](/docs/optimization/testing/interactive-testing) to validate your AI Agent's behavior. The test chat simulates a real conversation, allowing you to verify responses, test different user scenarios, and confirm that your configuration works as expected. With Interactive Testing, you can: * Chat with your AI Agent as an end user would * Test over voice with a browser call or a call to your phone * Simulate different user types by setting variables * Preview changes before exposing them to end users * Share test URLs with stakeholders for feedback ## Use cases Interactive Testing supports the following workflows: * **Configuration validation**: After setting up [Knowledge](/docs/knowledge), [Actions](/docs/automation/tools/api-tools/api-tool-control), or [Processes](/docs/automation/processes), test the AI Agent to confirm it responds correctly. * **User-specific testing**: Simulate different end-user segments by setting variables to verify personalized responses. * **Pre-launch review**: Share the test URL with stakeholders to gather feedback before going live. * **Change preview**: Restrict changes to test sessions using the `test_user` variable to validate updates before publishing. ## Capabilities & configuration Interactive Testing offers several capabilities for testing conversations and can be configured to match different scenarios. ### Test chat options The test chat provides options for flexible testing across different contexts. To open the test chat in a new window or another browser, click the **More options** icon ![](/_fern-img/3ee526b36308681790cf2406212393c20796c930949d49d91007d2664de07c2b.webp). * **Open in a new window**: Opens the test chat in a separate browser window. * **Copy URL**: Copies the test URL to your clipboard. You can use this URL to test your AI Agent in any browser window, even if you're not logged into Ada. > **Tip** > > Before you launch your AI Agent, share the test URL and get feedback from stakeholders across your organization! ### Voice testing If your AI Agent has Voice enabled and a voice phone number configured, the **Test AI Agent** button offers **Voice** alongside **Chat**. Voice testing supports two ways to connect: * **Web call**: the call runs in your browser, using your microphone and speakers. No phone required. * **Phone**: your AI Agent calls a phone number you enter. Select a language for the test before connecting. The list contains the languages enabled for your AI Agent, and defaults to the language configured for the number being called. Nothing connects until you click **Call**. Your last-used connection method, phone number, and language are remembered for the next test. ### Test variables If you've tailored your AI Agent's content to different users based on [Variables](/docs/automation/variables), you can set those variables in the test chat to simulate different end-user segments. > **Note** > > You can't simulate user types using sensitive variables. ### Chat persistence The test chat follows your configured [persistence settings](/docs/channels/chat/chat-configuration/data-controls#persistence-settings). When you reset the chat, it behaves according to these settings. ## Quick start Get started with Interactive Testing in just a few steps. For detailed instructions, see [Implementation & usage](#implementation--usage). #### Open the test chat From the lower left-side panel of your Ada dashboard, click **Test**. The Ada Web Chat launches. #### Send a message Type a message in the chat and press Enter to see how your AI Agent responds. #### Validate the response Review the AI Agent's response to confirm it meets your expectations. Continue testing with different messages and inquiries. ## Implementation & usage Start Interactive Testing, preview changes, simulate user types, and manage test sessions. ### Get started with Interactive Testing\[#start-testing] Once you've [connected your knowledge base](/docs/knowledge/content-ingestion/knowledge-integration) to your AI Agent, you can begin Interactive Testing. **To start a test session:** 1. From the lower left-side panel of your Ada dashboard, click **Test**. The Ada Web Chat launches. 2. Type a message in the chat and press Enter to see how your AI Agent responds. 3. Continue to test your AI Agent with various messages and inquiries. ### Test over voice\[#test-over-voice] **To start a voice test:** 1. Click **Test AI Agent**, then select **Voice**. 2. Choose **Web call** or **Phone**. 3. For a phone test, enter the number to call in international format, for example `+16135550183`. 4. Select the language for the test. 5. Click **Call**. During a web call you can mute your microphone, change your microphone and speaker in **Audio settings**, and end the call with **Hang up**. A phone call is controlled from your handset, so those controls aren't shown. While the call is connecting, a link to the conversation appears so you can follow the transcript as it happens. The link stays available after the call ends. > **Note** > > Voice test calls are recorded as test conversations, so they're excluded from your reporting and are subject to the same `test_user` availability rules described in [Test changes before they go live](#test-changes-before-they-go-live). On a web call the end user appears as **Anonymous**, since a browser call has no phone number. ### Test changes before they go live You can preview a change before exposing it to end users by restricting the change's availability to test sessions. **To test a change before publishing:** 1. Create or open the change you want to test. 2. Click **Set availability → Based on the following rules…** 3. For the **Variable** field, search for and select **`test_user`**. 4. Switch the comparator from **Equals (=)** to **Is set**. 5. Save the change. It now applies only to sessions flagged with `test_user`. 6. In the dashboard's lower-left corner, click the purple **Test AI Agent** button to open a chat. 7. Chat with the AI Agent to verify the new behavior. 8. When you're satisfied, reopen the change, switch availability to **Everyone**, and click **Save** to publish it for end users. ### Impersonate different user types \[#impersonate-user-types] You can set variables in the test chat to simulate different end-user segments and verify that responses are personalized correctly. **To impersonate a user type:** 1. In the test chat, click the **User** icon beside the top of the test chat. ![](/_fern-img/a485856100b4981713905b88a5a4224debedba8d6d10fd512265066dd3f7f81a.webp) 2. In the **Set test variables** window, set the variables corresponding to the user type you'd like to test, and start chatting with your AI Agent. If you start typing a variable value, you can see a list of values from previous conversation data. For example, if you're using the `region` variable, any values stored in `region` from previous conversations will appear as options. You can click one of these options to select it, or continue typing to enter a different value. ![](/_fern-img/17948c6eb0e758ac5e4cec645bfd12f2b64d3d670b02ba814882ba93135d12f2.webp) ### Reset or end your test chat \[#reset-end-chat] You can reset the test chat to start a new session or exit when finished. **To reset the test chat:** 1. Click the **Refresh** icon. The chat resets according to your configured persistence settings. **To end the test chat:** 1. Click anywhere outside the test chat to exit and return to your dashboard. ## Related features * **[Simulations](/docs/optimization/testing/simulations)**: Create and run automated test cases with defined expected outcomes to validate changes at scale before they affect end users. --- Have any questions? Contact your Ada team, or email us at [](mailto:help@ada.cx?subject=Help%20Docs%20inquiry).