Before you start
Confirm that the relevant entry is available in your workspace. You need permission to deploy an agent and access to the connections it requires. For a self-hosted worker, also prepare the runtime and network access described by its deployment instructions.Inspect the entry
Read the worker’s job, deployment options, required model access, external integrations, and limitations. Decide whether its behavior matches your intended workflow. Names alone do not establish permissions or behavior. For example, the catalog’s Remediation worker proposes fixes and never changes the live system. The Kubernetes RCA entry can requirepods/exec
for diagnostic skills. Review the implementation’s requirements before approving access.
Configure and deploy
Follow the Worker catalog procedure to select a unique agent identity and specific active connections, review the configuration, and deploy. Availability can differ between workspaces, and not every example agent is a marketplace or catalog entry. The server pins a catalog deployment to an exact image tag. Do not assume a listing implies automatic adoption of all future upstream changes. Treat the deployed agent’s configuration, access, and validation as account responsibilities.Verify the result
Confirm the agent is online, run a representative task, and inspect its evidence. Check that it used the intended external account, returned the expected result, and stayed within the access you configured. If an entry is absent or unavailable, check account availability rather than substituting a similarly named worker with different behavior. If deployment reports missing credentials, correct the required provider connection or model access and retry.What you’ve built
You have selected a packaged implementation and established a tested agent configuration in your own account.Next steps
Worker catalog
current entries, requirements, and deployment API
Agent examples
patterns for a custom implementation
Integrations
provider authentication and connection lifecycle