The agent can list repositories, tags, and images under your namespace; view organization details, teams, and members; manage webhooks; and retrieve image metadata like digests and platform support. Access depends on the scopes assigned to your PAT. Read-only tokens allow lookups; read-write tokens enable creating repos and managing team membership.

No. The agent interacts with Docker Hub's REST API for metadata operations: listing repos, checking tags, managing teams, and configuring webhooks. Pulling and pushing images requires a Docker daemon, which is outside the scope of this integration. The agent helps you find the right image, but the actual pull happens in your terminal or CI pipeline.

The agent respects Docker Hub's visibility settings. If your PAT has access to a private repository, the agent can list its tags and images. Public repositories are accessible regardless. When creating a new repository, you specify the visibility, and the agent applies it. No private data leaks into conversations about public repos.

Your PAT is stored encrypted in Tars' tool configuration and used only to authenticate API requests. Repository metadata, tag lists, and team information are fetched live from Docker Hub during each conversation. No copies of your registry data are maintained. You can rotate your PAT at any time from Docker Hub's security settings.

Yes. If your PAT has access to multiple organizations, the agent can list them and query repositories, teams, and members across any of them. Specify which organization you are asking about in the conversation, and the agent targets the correct namespace.

The web UI requires logging in, navigating to the right namespace, clicking through repository pages, and filtering tags manually. The agent answers in one message. Ask 'What are the latest tags for our api-service repo?' and get the answer immediately, without leaving your current workflow in Slack, Teams, or WhatsApp.

The agent supports deletion operations if your PAT allows it. However, you can configure guardrails in Tars to require confirmation before destructive actions. For example, the agent can ask 'Are you sure you want to delete the staging-api repository? This cannot be undone.' before proceeding. Or restrict deletion to admin-approved workflows only.

The agent handles this cleanly. If a repository or tag is not found, it reports that no matching resource exists and suggests checking the namespace or tag name for typos. It does not throw errors or display raw API responses. Users get a helpful message explaining what went wrong.