Skip to main content

Calling an Agent from chat

Utilize your custom Agents directly from the Work page.

Updated this week

On the Work page, you can call any active agent directly from the chat input β€” no need to navigate to the Agents section. Just type @ followed by the agent's name, and Coworker will run that agent in the context of your current conversation.


How to call an Agent from chat

  1. Go to the Work page
    Navigate to app.coworker.ai and open a new or existing chat.

  2. Type @ in the chat input
    A list of your active agents will appear. You can scroll through the list or continue typing the agent's name to filter it.

  3. Select the agent
    Click the agent you want to run, or finish typing its name and press Enter to select it.

  4. Add any additional context (optional)
    After selecting the agent, you can include extra instructions or context in your message before sending. The agent will use this alongside its configured instructions.

  5. Send your message
    The selected agent will run and return its response directly in the chat.

Only agents that are active and saved in your Agent Builder will appear in the @ list. If you don't see an agent, make sure it has been created and enabled in Agent Builder.


What happens when you call an agent

When you call an agent from chat, it runs with the tools and data sources it was configured with in Agent Builder β€” not the full context of a standard Deep Work session. This means:

  • The agent's output is scoped to the data sources it has access to

  • If the agent has an output delivery configured (e.g., a Slack channel), that delivery will also trigger

  • Starting a new chat after calling an agent returns you to the standard full company context


When to use this vs. running an agent from the Agents page

Calling via @ in Work

Running from the Agents page

Quick, on-demand usage within an active conversation

Running on a schedule or reviewing execution history

Useful when you want to add context or follow-up in the same chat

Best for automated or recurring workflows

Ideal for ad-hoc use of an agent mid-task

Best for managing or testing agent configurations

Did this answer your question?