---
title: "Managed Agents: Testing, Publishing, Unpublishing and the Activity Feed"
description: "Test a Managed Agent safely, publish and unpublish it, read the Activity feed, run parallel conversations, and copy an agent within your account."
url: https://help.aesthetixcrm.com/en/articles/2229-managed-agents-testing-publishing-unpublishing-and-the-activity-feed
collection: "Agent Studio"
updated: 2026-11-20
language: en
---

# Managed Agents: Testing, Publishing, Unpublishing and the Activity Feed

Test a Managed Agent safely, publish and unpublish it, read the Activity feed, run parallel conversations, and copy an agent within your account.

Every Managed Agent follows a draft, test and publish lifecycle. You manage all of it in **AI Agents > Agent Studio > Managed Agents**: open an agent to use its testing panel, the publish control, and the **Activity** tab. This guide covers testing, publishing, unpublishing, reading the Activity feed, running parallel conversations, and what carries over when you copy an agent's setup inside your account.

> **Protect patient privacy.** Aesthetix CRM's AI features are not approved for Protected Health Information (PHI). Do not put PHI into an agent's instructions, files, memory, knowledge or prompts, and never use real patient records while testing. Test with your own contact details or a clearly fake test contact.

## Test Before Publishing

Testing lets you confirm the agent's responses, actions and boundaries before it touches live activity.

1.  Open the Managed Agent and open its **testing panel**.
    
2.  Run realistic requests that exercise its Knowledge Base, CRM actions and instructions.
    
3.  For responses that use web search, review the **Sources** and open the citations to check the information.
    
4.  If media generation is enabled (Image, Audio or Video generation under **Capabilities**), test a sample asset.
    
5.  Refine the agent with the chat-based editing sidebar and repeat the test.
    
6.  For agents with triggers, simulate the trigger event and confirm it starts the agent and passes the expected data. See the Triggers Reference guide.
    
7.  Check the **Activity** tab to see exactly what the test run did.
    

Important testing notes:

-   **Tests can send real messages and perform real actions**, including changing CRM records. Always test with your own contact information.
    
-   For scheduled follow-ups, **Test Agent** skips the waiting periods so you can check the sequence without waiting hours or days, but the actions are still real.
    
-   You can test and chat with an agent even while it is unpublished.
    
-   Define explicit boundaries in the agent's instructions, such as requiring human approval for suggested changes, and test that those boundaries hold.
    
-   If your agent uses a Knowledge Base, use the Knowledge Base Retrieval Tester to check that important content can be retrieved.
    
-   If the agent books appointments, remember that AI booking works only with native Aesthetix CRM calendars, not EMR calendars. Test against a native test calendar.
    

## Publish a Managed Agent

Publishing makes the configured agent available for live events. After testing, change its status from **Draft** to **Published**.

1.  Review the agent's instructions, triggers, tools, Knowledge Base access and capabilities.
    
2.  Click **Publish** in the upper-right area of the builder.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-9Vr8v8neZn9HUjxbJ85tXVjB9CsbxSBjg91da74gBuDc0ztblhyNPcyIjuRgOdZ33PLxruti5c9.png?v=1&k=us1&s=api&e=1795348800&sig=6JZIZVyupMsXJkeR3HwcvKuZWLaLfCggzJQr1sxBCks)
3.  Confirm the agent now shows as published.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-RJLofdEjdoCmHzkLtOtffPafaUQjXah3DGdA8vYOwH1jIQ8pNcUNo76VeA2Qh3RDpw3L8EdJG2f.png?v=1&k=us1&s=api&e=1795348800&sig=PAyemfnqxrKAHGvsh9HnDJGJy0HPsCntoxdBFBa7J7I)

What publishing does:

-   Activates the agent's production triggers (schedules, webhooks, appointment, CRM, form, tag, survey and other triggers).
    
-   Makes the API and workflow invocation options use this version.
    
-   When the agent uses Facebook lead form or survey triggers, checks that the selected Page, lead forms or surveys still exist and flags invalid selections.
    
-   New runs follow the latest published configuration. If you edit later, republish for changes to reach production.
    

## Unpublish a Managed Agent

Unpublishing temporarily takes the agent's automated triggers out of production without deleting anything.

**What happens when you unpublish:**

-   Production triggers stop running.
    
-   Armed scheduled runs are cleared, so the agent stays inactive in production.
    
-   Webhook URLs stop accepting requests, and API calls to the agent return `404`.
    
-   Pending scheduled follow-up runs stop.
    
-   An event already in progress appears as **Skipped** in Activity history with a reason.
    
-   Your configuration stays intact: triggers, apps, knowledge and instructions.
    
-   You can keep editing, testing and chatting with the agent.
    

**To unpublish:**

1.  Go to **AI Agents > Agent Studio** and open the published Managed Agent.
    
2.  Click the **Edit Agent** icon next to Publish in the upper-right corner of the builder.
    
3.  Click the **three-dot** icon next to Publish and select **Unpublish**.
    

To return it to production, click **Publish**. The latest saved configuration is published and the triggers become active again.

## Activity Feed

The Activity feed is a unified timeline inside each Managed Agent. It shows live chat conversations and automated trigger runs together, newest first. It is specific to each agent, so open each agent's **Activity** tab to review its runs.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-MDqzQzeuDYYIwwLFlEm2UJh0GaHoLLNqsI2pTuDichaogIGCEpH1OOO5ywdHNN6duuklgYE6Ftz.png?v=1&k=us1&s=api&e=1795348800&sig=WYhDIPENyNmeEMUTl1Z8dUQb8Vh51MX2SaVEQfcJpts)

### Open it

1.  Go to **AI Agents > Agent Studio > Managed Agents**.
    
2.  Select the agent.
    
3.  Select the **Activity** tab.
    
4.  Open a run or conversation to see its execution log.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-HqdaY7EvSOGzs0IjOpNv6FmVHb06RWmgD15sTgLTkp22fRLgrgeeyQgUifLAGICSJ1UuWdPOXMs.png?v=1&k=us1&s=api&e=1795348800&sig=XyDD0DwdviHO-AujqU0VXLuSUis4BHdfrRH5Qb30DLc)

### What each row shows

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Element</strong></p></th><th colspan="1" rowspan="1"><p><strong>Meaning</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Run icon</p></td><td colspan="1" rowspan="1"><p>The type of activity. Automated trigger runs can use a lightning-bolt icon, live conversations a chat icon, and appointment runs a calendar icon.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Run status</p></td><td colspan="1" rowspan="1"><p>Completed, failed, or still in progress. Follow-up runs can also show Waiting or Skipped.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Timestamp</p></td><td colspan="1" rowspan="1"><p>When the run started, useful for matching it to a submission or event</p></td></tr><tr><td colspan="1" rowspan="1"><p>Trigger type</p></td><td colspan="1" rowspan="1"><p>Whether the run came from a trigger or a conversation, such as an Appointment tag, <strong>Facebook Lead Form Submitted</strong>, <strong>Survey Submitted</strong> or <strong>Webhook received</strong></p></td></tr><tr><td colspan="1" rowspan="1"><p>Context</p></td><td colspan="1" rowspan="1"><p>For automated runs, the form, tag event, survey or other event that started the agent. For chats, a preview of the person's message.</p></td></tr></tbody></table>

Runs that appear include form submissions, tag changes, Facebook lead form submissions, survey submissions, appointment events, scheduled runs, webhook runs, API runs and other supported trigger events.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-3CgMdhelVymeEiZcC3F9uggbT2isNldEnMCqUERxcmpMrD7dUuw5KgPQDgGAAGSUlJUcjGGTcqJ.png?v=1&k=us1&s=api&e=1795348800&sig=4l2Su8BYmcIEX5USfZqqlyS2Vk1znRAslqbDNn_QaeM)

### Execution log

Opening a run shows:

-   **Agent Conversation View:** the conversation and the agent's replies for that run.
    
-   **Execution timeline:** the steps the agent completed in order, with timing and token usage where available, and which skills or tools it used. Use it to find where a run went differently than you expected.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-H3UGopIJ7Gla7E4YEpQe6AyR4LXqFjGE8Y6G8HsYkMQkbaUSHUd4Rarw4SXlya77XNiOWHgjgyU.png?v=1&k=us1&s=api&e=1795348800&sig=04X-rrh8JtZM9EF5ZM-Sith85nAQDG64z8Hqz9EKEXc)
-   **Controls and metadata** at the top: the conversation title, refresh, timing metrics, token usage, **Copy ID**, and **View in Agent Logs**.
    

For step-level tracing, metrics and usage across agents, see the Agent Logs, Metrics, Monitors and AI Usage guide.

### Rate agent replies

In the Agent Conversation View you can rate individual agent replies.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-WrMTPERsBA3A03rqWicYUdh9RH0OFa1B0SabzstH52YcNyFYKVVewMD861xcZqUYEUAtpLPLbSD.png?v=1&k=us1&s=api&e=1795348800&sig=gabOp59aZL6xiYRV12zkhVMWNqLeDnseP2Nn_O9peGE)

-   **Thumbs up:** the reply was useful, accurate and appropriate.
    
-   **Thumbs down:** the reply had wrong information, missed context, took an inappropriate action, or was off-base. A short follow-up opens where you can explain what went wrong, and you can add an optional note.
    

Feedback is attached to the specific reply you rated. After you submit, the rating stays visible when you return, previously rated replies do not appear unrated again, and you can reopen a rating to review what you submitted. Use the examples to improve instructions, knowledge coverage and overall behavior.

### Waiting runs

Follow-ups that are paused show as **Waiting** with their resume time. Open the run to see steps or click **Cancel wait** to stop it.

### Debugging a failed run

1.  Open the failed run in Activity.
    
2.  Review the Agent Conversation View and execution timeline to find where the problem started.
    
3.  If the agent did not run at all, confirm it is published and that trigger filters match.
    
4.  Adjust instructions, retest, and publish again.
    

## Parallel Conversations

You can work on several independent tasks with the same agent at once. Each conversation is its own workspace.

-   **Up to 10 conversations can run at the same time.**
    
-   **The 20 most recent conversations** stay readily available for quick return. This is a separate limit from the 10 active ones: one is about simultaneous processing, the other is about recent availability.
    
-   **Background processing:** when you switch to another conversation, the first keeps working. Reopen it to return to its live stream.
    
-   **Independent state:** each conversation keeps its own messages and history, scroll position, loading state and error state, so nothing mixes together.
    
-   **Conversation-specific composer:** the message box belongs to the conversation you are viewing, so you can keep sending requests while others process.
    
-   **Monitoring:** the **Activity** tab beside the chat shows running work with a pulsing in-progress indicator, and outcomes appear when runs settle. Running conversations get their own headings.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-FpHmfOWTQkPEeaUZ2Ljfw7kuP2dqpxg9Lonsw7Lp8FHrrzW3IdPwt8pNsYAdptxehakvSG0KFuS.png?v=1&k=us1&s=api&e=1795348800&sig=UkYVGbYCYGRnwfA3vkqFvFsF7-8hTJu6TkcdEmOFizM)
-   **Consistent media:** agent chat, the builder test panel and preview use the same media display.
    

Examples of separate conversations: drafting a newsletter, pulling information for a report, working through a contact, preparing campaign content.

To use it: open the agent, send your first request, open another conversation while the first runs, send the next request, repeat up to 10, and use **Activity** to follow progress. This is separate from Ask AI multi-threading, which has its own experience.

## Cloning and Sharing a Managed Agent

You can reuse a proven agent by copying its setup inside your own account, for example a consult-follow-up agent you want to adapt for a second service line. For the step-by-step process of duplicating and organizing agents, see the Agent Studio: Folders, Cloning and Sharing Agents guide. This section explains what carries over and what to double-check.

**What a copy carries:**

-   Agent flow structure and node configuration
    
-   Prompts and instructions
    
-   Triggers
    
-   Skills and capabilities
    
-   Connected Knowledge Bases, when the Knowledge Base itself is available to the copy
    
-   Version state, such as Draft, Staging or Production
    
-   Other supporting items the agent depends on, such as workflows or Custom Values, which need to exist for the copy to work
    

**What does not carry:**

-   Live account activity and customer records
    
-   Integrations, third-party connections, credentials and other account-specific resources, which usually need to be set up again
    

**Overlaps and conflicts:** before copying an agent into a place that may already contain a related agent or an earlier version, review the overlap so you do not accidentally replace a working configuration. Keep the original until the copy is verified.

**Verify the copy:**

1.  Go to **AI Agents > Agent Studio > Managed Agents** and confirm the agent appears.
    
2.  Open it and review its name and description, triggers, skills, capabilities, Knowledge Base configuration, instructions and required integrations.
    
3.  Missing Knowledge Base content usually means the Knowledge Base was not included.
    
4.  Test it, then publish when ready. A copy does not necessarily run correctly until connections and permissions are in place.
    

## Frequently Asked Questions

**Does unpublishing delete my Managed Agent?**

No. It pauses production triggers and keeps the configuration. Republish to bring it back.

**Can I still test or chat with an unpublished agent?**

Yes. The testing panel and chat stay available so you can validate your draft before publishing again.

**What happens to a scheduled or waiting run when I unpublish?**

Armed scheduled runs are cleared and pending follow-up runs stop. An event already in progress shows as Skipped with a reason.

**Are 10 parallel conversations and 20 recent conversations the same limit?**

No. Up to 10 can process at once, and the 20 most recent stay ready for quick access.

**Does switching to another conversation stop the first one?**

No. It keeps running in the background and you rejoin its live stream when you reopen it.

**Can I see all my agents in one Activity feed?**

No. The feed is per agent. Open each agent's Activity tab, or use Agent Logs for broader review.

**What do the timing and token metrics mean?**

Timing shows how long a step or run took. Token usage shows how many tokens supported steps or the whole run consumed.

**Can testing affect real contacts?**

Yes. Tests can send real messages and make real CRM changes, so use your own contact details or a clearly fake test contact.

**Can I use patient records to test an agent?**

No. Aesthetix CRM's AI features are not approved for PHI. Keep PHI out of instructions, files, memory, knowledge, prompts and test conversations.

**Will a copied agent work immediately?**

Not necessarily. Connections, permissions and missing supporting items may need setup before it behaves as intended, so review and test it first.
