Help Center
Start with the request ID, key policy, model catalog, and usage record.
Troubleshooting and common questions
Why do I get “Invalid API key”?
Confirm the entire ori_… value is present, the key is active and unexpired, and your shell did not keep an old value after rotation. Create a new key if the raw secret is lost.
Why is a model missing or rejected?
Call GET /v1/models with the same key. The model must be active, permitted for the key, available, and support the requested capability.
Why did the request hit a limit?
Check the key's RPM, TPM, concurrency, token, and cost limits, plus the available account balance. Inspect RateLimit-*, Retry-After, and the usage dashboard before retrying.
Why does Claude Code ignore my base URL?
Run /status. User, project, or editor settings may override shell exports. The base URL must be exactly https://origamiai.my, without /v1.
Why can’t a browser call the API?
Browser access may be restricted to approved origins. Contact your administrator if your application is blocked. Do not ship a long-lived virtual key in client-side code; proxy calls through your backend.
What should I send to an administrator?
Send the time, endpoint, public model, HTTP status, and X-Request-ID. Never send the raw virtual key, full sensitive prompt, or a secret environment file.
Contact support
Reach a human with the request context needed to help quickly.
Platform use and data policy
These guidelines apply when accessing Origami and supplement any agreement with your service administrator.
Authorized use
Use the service only within your organization's approved purpose, model access, content rules, applicable law, and service terms. Do not probe, bypass, overload, or resell access without authorization.
Credentials
You are responsible for securing virtual keys and promptly revoking or rotating exposed credentials. Activity performed with an active key is attributed to its owner and scope.
Data handling
Prompts and outputs are sent to the model service used to complete the request. Origami may retain request metadata and usage needed for security, accounting, support, and service operation.
Billing and budgets
Dashboard amounts are calculated from applicable pricing and recorded usage. Limits reduce risk but remain subject to processing delays and estimated usage.
Availability
Requests depend on network connectivity, model availability, and account quota. Models or features may change, and asynchronous jobs may expire or fail.
Enforcement and changes
Administrators may suspend accounts, keys, or model access to protect the service, enforce policy, or comply with legal requirements.