The agent uses natural language understanding to match your request to the correct API endpoint. If you say 'create an issue in the frontend repo,' it maps to the create issue endpoint with the right owner and repository parameters. Ambiguous requests get clarifying questions before execution.

Tars requests repo, user, gist, notifications, project, workflow, and codespace scopes. These cover the full range of GitHub operations the agent can perform. You authorize through GitHub's standard OAuth flow and can revoke access from your GitHub settings at any time.

Yes. The agent can trigger workflow dispatches, list workflow runs, check run status, download logs, cancel in-progress runs, and re-run failed jobs. Your CI/CD pipeline becomes fully manageable through conversation, from triggering deployments to investigating failures.

No. Tars calls GitHub APIs in real time and does not maintain copies of your source code. File contents fetched during conversation are used to formulate responses and are not persisted. Your code remains exclusively on GitHub under your access controls.

Yes. The OAuth token grants access to all organizations and repositories the authenticated user can access. The agent works across personal repos, organization repos, and even repos where you are an outside collaborator, all in one conversation.

GitHub Copilot Chat helps with code writing and explanation inside your IDE. Tars AI Agent manages GitHub operations: creating issues, reviewing PRs, triggering workflows, managing permissions, and querying across repos. It is an operational tool, not a code generation tool, and works in any channel including Slack and WhatsApp.

Access is governed by your GitHub OAuth token permissions. If you authorize with a user who only has access to certain repos, the agent inherits those restrictions. For finer control, create a dedicated GitHub user or machine account with access limited to specific repositories.

GitHub allows 5,000 authenticated requests per hour. If the rate limit is approached, the agent informs you of remaining capacity. If exceeded, it waits for the reset window and reports the delay. For high-volume teams, GitHub Apps with higher rate limits can be configured.