Concepts

The models behind the features. Read these when something surprised you and you want to be able to predict it next time, rather than look it up again.

Pages

• AI agents and the MCP server — what an AI assistant can read through the MCP server, and how to connect one
• Token scopes — the one scope every operation needs, bundles, the scopes no bundle carries, and binding a token to one place
• Audit — which operations are recorded, what the tamper-evident chain proves, and who can read which log
• Quotas — the caps on what you create, the 429 at a cap, enforced and deferred caps, and asking for more
• Egress — the byte budgets on what you pull, which operations are metered, and when each is charged
• Caching — how long you may keep a response, and how an ETag makes checking again cheap
• Listing isolation — why a listing shows only what you can see, and why a private resource answers 404
• Pagination — reading a listing to its end with next_token
• Live streams — watching a merge request, a run or a log as it changes, and resuming a dropped stream
• Dry runs — previewing a change with dry_run=true before you make it
• Errors — the error envelope, code and reason, and which failures are worth retrying

Related