Deploying the agent

When you decide that an agent is ready for use by your team, you need to deploy it to move it from draft to live.

An agent in draft is under development and not yet available in channels like chat. Deploying the agent makes it live with the current set of functions that are used to ground its behavior.

After deployment, you can use the agent in the channels set up during its development. For more information, see Connecting to channels.

Before you begin

Make sure that your agent includes the required details, such as a name and description, before you deploy.

Deploying your agent

Click Deploy from the agent configuration page to view the Pre-deployment summary page. It includes a summary of its components, connection status, and deployment controls.

Deploy button
Figure 1. The Deploy button.

Pre-deployment summary configuration

This part shows the agent's key components in one place. You can review the selected model, uploaded files, linked tools, defined instructions, and any collaborators before you deploy.

Section Information displayed
Profile Agent description and selected model
Knowledge Uploaded district files
Tools Associated tools and their connection status
Collaborators List of collaborators (if any)
Behavior Instructions expected by the agent

A green checkmark Supports appears next to each tool when its security schema matches the linked connection.

Connections

This part outlines the structure of the agent's active connections. Each column describes how the agent interacts with external systems and tools.

Heading Description
Server A target server or an endpoint that is linked to the connection.
Associated tools Tools that reply in this connection. Multiple tools can share one connection.
Live Indicates whether the connection is active in the production environment.
Connection status It shows whether the connection is established or broken.
Note: The Deploy button is disabled if any connection that is associated with the required tool is not connected. Reconnect the tool by using the Setting up a connection.

Connection validation

The deployment interface displays only connections to the live environment. It helps ensure that the agent is linked to production-ready systems and avoids confusion with draft configurations.

If a required tool is not connected, the Deploy button remains disabled. To continue, set up the missing connection by using the following steps.

Setting up a connection

Follow this procedure to configure a new connection for your agent, including authentication and credential settings.

  1. Click Add connection +.

  2. Enter the Connection ID.

  3. Enter the Display name.

  4. Click Save and continue.

  5. To enable single sign-on, toggle Single sign-on (SSO) on.

  6. Select an Authentication type:

    • API Key
    • Basic Auth
    • Bearer Token
    • Key Value Pair (only for Team credentials)

  7. Enter the required details for the selected authentication type.

  8. Choose a credential type:

    • Member credentials: Each user can access the application by using their own credentials.
    • Team credentials: All users access the application by using shared team credentials.
Note: Use team credentials when centralized access is preferred. Use member credentials when individual authentication is required.

  1. Click Add connection.

    The connection appears in the table. To edit it, click the pencil icon Pencil icon.

  2. Click Deploy to complete the setup. You return to the agent configuration page.

After setup, the agent's status is indicated by a live icon Live icon. Hover over the icon to see the deployment timestamp in your local time zone.

Undeploying the agent

The undeploy option is available when the agent is live. Selecting it opens a confirmation dialog that outlines the impact:

  • Production instances are interrupted.
  • Access to the live environment is restricted.
  • The agent becomes unavailable to users.

After undeployment:

  • The agent returns to the draft state.
  • The interface updates immediately.
  • The agent is removed from live search results.

Agents in the draft state do not appear in chat search results.

To undeploy an agent,

  1. Go to the agent configuration page.
  2. Click the overflow menu Overflow menu icon.
  3. Select Undeploy.
  4. Click Undeploy to confirm.

What to do next

After you deploy the agent, your agent becomes available across all configured channels. For more information on using agents in watsonx Orchestrate chat, see Using agents in the watsonx Orchestrate chat.