Agent activation after production deployment
Key Takeaway
Deploying an agent to production creates a new version there, but that version doesn't go live until someone explicitly activates it.
Moving an agent from a sandbox to production takes two separate steps. Deployment copies the agent's metadata into the target org, but it doesn't put the agent in front of users. Activation is the step that makes a specific agent version live on its connected channels. Keeping the two steps apart lets you check the deployed agent and decide exactly when it goes live.
Deployment and activation are separate steps
Agents are defined as metadata, so they move between orgs with standard Salesforce deployment methods, such as change sets, the Metadata API, or Agentforce DX.
A deployment only places the agent's metadata in the target org. It creates a new version of the agent there. The new version doesn't go live until someone explicitly activates it, so you have full control over when you release it.
Activation is the step that makes the agent operational and available to users. If the agent is already connected to one or more channels, including the Agentforce panel, activating it makes the agent immediately available to users on those channels.
Because activation is separate from deployment, you can also do a controlled rollout. Salesforce recommends starting with a pilot program: a small, controlled group such as one team. Then, after refining the agent's performance, you gradually roll it out to other users.