GetVibeAPIFrequently asked questions
Answers to the questions our support team sees most often: models and channels, errors, pricing, media jobs, API keys, billing and privacy.
Usage
Models, channels, errors, pricing, caching and media generation.
How do I choose a model and a channel?
Send the model id from the catalog, for example gpt-6-luna. A model can be available through several channels; pin one with a colon suffix, for example gpt-6-luna:cx. Without a suffix the router picks an available channel and can move to the next one on failure. With a suffix the channel is pinned. GET /v1/models lists the ids your key can call.
What does the suffix after the colon in a model name mean, and what should I do on a 503?
It is the model's channel: the same model from another source, with its own price, speed and cache support. Without a suffix the request picks an available channel and moves to the next one on failure. With a suffix the channel is pinned and does not switch on errors, so on a 503 drop the suffix or pick another channel on the model's card in Models & pricing. How a key switches channels is set by its routing strategy in the key settings.
How is cached input priced?
A cache read is usually billed at 0.1× the input price; some models have a different coefficient. The exact cached-input price and coefficient are shown on the model page and in /api/catalog. Usage responses carry cache_read_input_tokens and cached_tokens.
What are min_price and max_price?
They are optional price bounds in USD per 1M tokens, compared with our published sell price. max_price is a ceiling: a request is never charged above it. min_price is a floor that excludes cheaper channels. If no channel fits, the API answers 422 price_bound_no_lane with the channel list. Both keys are removed before dispatch and never sent upstream.
How do media generations and their pricing work?
Get a quote first with POST /api/media/price (or read the price in the media catalog at /models), then create the job with POST /api/media/generate, which returns 202 and a jobId. Poll GET /api/media/jobs/{jobId} until success or fail, or pass webhook_url. A successful job is charged the quote; a failed job is not charged. Results are served from our domain for 7 days (resultsExpireAt), then answer 410.
Where are the setup guides for Claude Code, Codex, Cursor and OpenCode?
See /docs/integrations for the overview and /docs/claude-code, /docs/codex-cli, /docs/cursor and /docs/opencode for per-tool setup. Each guide shows the base URL, environment variables and key setup.
Cabinet
Top-ups, keys, two-factor authentication and receipts.
The confirmation code does not arrive by e-mail — what should I do?
Check your Spam and Promotions folders. You can request a new code once a minute; always enter the code from the latest e-mail. If the address has a typo or the e-mail never arrives, write to Support with the correct address.
I paid but my balance did not go up — what should I do?
The balance is normally credited as soon as the payment is confirmed. Open Billing → Payment history: if the payment is still Pending although the money has left your account, write to Support and attach the receipt or transaction ID. We will check the payment and credit it manually.
How do I top up and what is the minimum?
Top up in Billing. The top-up form lists the payment methods enabled for your account — the card rails while they are switched on, and crypto through NOWPayments or CryptoBot when that rail is enabled. The card-rail minimum is currently $1; the exact minimum is shown on the form.
What can I configure on an API key?
A key can carry a spend limit, an allowed-model list, an IP allowlist and a routing strategy: Default (the first available channel), Only default (no switching to alternatives) or In order (failover).
How do I enable two-factor authentication?
Open Settings → Two-factor authentication and enable an authenticator app (TOTP) or an email code. The second factor protects sign-in and payment-detail changes.
Where do I get a receipt or an invoice?
Payments are processed by the payment provider, which issues the receipt for the charge. For company billing details or a document request, write to [email protected] and include the payment date and amount.
Privacy
What happens to the content you send to a model.
Where is my request content processed?
Request content — text, and any images, audio or video you attach — is transmitted to the model providers that serve your request. Those providers may process it outside your country.
Do you keep my prompts or train on them?
No. We do not keep request content as chat history and do not use it to train models. Billing and security metadata (key identifiers, token counts, payments and balance) are stored separately.
How long are files and results kept?
Inline base64 inputs are removed after 24 hours. Files uploaded through POST /v1/files are kept for 7 days. Generated results are served from our domain for 7 days after the job completes (resultsExpireAt), then deleted.
Useful sections
Errors and limits
Response codes and what to do about transient failures.
Tool setup guides
Claude Code, Codex, Cursor, OpenCode and other clients.
Claude Code
Anthropic CLI setup through /v1/messages.
Codex CLI
Responses API setup with wire_api = "responses".
Cursor
Override OpenAI Base URL and a custom model.
OpenCode
The opencode.json provider and config download.
Privacy policy
The full policy and how to contact us about data.
Terms of Service
Rules of use and ownership of generated results.
Support
Open a ticket if this page does not answer your question.
Still have a question?
Open a support ticket and we will help with your specific case.