Frequently asked questions about provod.ai
Short answers about starting, billing, API access, models, teams, and support.
Last updated on
This page gives short answers and links to complete instructions. Check current models, prices, payment methods, and operation states in the cabinet instead of relying on old numeric examples.
Getting started and account access
How do I start with provod.ai?
Create an account, choose a task, and, if you need the API, send one small non-streaming request. The complete path and runnable example are in Start with provod.ai.
How do I register or sign in?
Register by phone or use Yandex ID when that button is shown; when returning, choose the method already connected to the account. All sign-in paths and recovery boundaries are in Account and access.
What if the SMS code does not arrive?
Check the number and request a new code when the control becomes available; enter only the newest message. If the problem remains, send support the masked number, time, and error under the Account and access guidance.
What happened to my Promptra account?
Promptra is the former service name, and old website and cabinet links redirect to provod.ai. Sign in with the previous method and verify the workspace, balance, and masked key prefixes by following the account check.
Payment, balance, reservations, and usage
Is this a subscription or usage billing?
provod.ai uses prepaid metered billing: confirmed usage is charged against one shared RUB balance. Token categories and captured-charge checks are explained in Usage and costs.
How do I top up, and what is the minimum?
Open the active workspace's balance and use the payment method and minimum shown by the current form. An individual profile receives the displayed card and SBP controls, while an ООО or ИП profile creates an invoice; see Balance and top-ups.
Why is part of the balance reserved?
Before a request, the service temporarily holds an estimate, so available funds can be lower than total funds, especially during concurrent work. After settlement, confirmed usage is charged and the unused portion is released under Balance and top-ups.
Why did an agent cost more than the last message suggests?
An agent can make many separate requests and resend history and tool results in each one. Compare the complete request history in Usage as described in Usage and costs.
Can an interrupted request be billed?
A failure before confirmed output or usage has no usage charge, but an already delivered stream portion or confirmed usage can be billed. The settlement boundary and safe response are covered in Usage and costs and Limits and retries.
API keys and connection
How do I create and store an API key?
Create a key in the active workspace and save the complete value immediately because it is not shown again. Revoke and replace a lost or exposed key under API authentication.
Which base URL should I use?
OpenAI-compatible clients use https://api.provod.ai/v1; Claude Code receives https://api.provod.ai without /v1 because it appends the Messages path. Choose the exact setup from Coding tool integrations.
Can browser code call the API directly?
Do not put a platform key into frontend source or a public browser bundle where another person can extract it. Call provod.ai from a trusted backend as described in API authentication.
Is the Responses API supported?
Yes. POST /v1/responses supports the portable Responses flow, including streaming, function tools, previous_response_id, and Conversations. Hosted tools such as Responses web_search and file_search are not supported; see Responses and Conversations.
Errors and long requests
What does an API error mean?
Read the HTTP status and public error.code or top-level code before choosing an action from the table. Error shapes and the safe diagnostic checklist are in API errors.
Should I retry after 429 or 5xx?
API_KEY_SPEND_LIMIT_EXCEEDED requires waiting for resetAt or changing the budget, while a transient failure can be retried with bounds only before output. Honor Retry-After when present and follow Limits and retries.
How should I handle a long or interrupted stream?
Streaming delivers a long response incrementally but does not remove every timeout; a close before the terminal event means the result is incomplete. After the first output, preserve it and do not retry automatically, following Limits and retries.
Models and capabilities
Where can I find current models, prices, and limits?
Use the model catalog or GET /v1/models because catalog contents and availability change. Copy the exact ID and inspect capabilities under Choose a model.
Why does a model show a lock?
In a personal workspace, FIRST_TOP_UP_REQUIRED means the model requires the first real top-up; organizations can follow different rules. Check the current reason and active workspace under Choose a model.
Are image generation and editing available?
Yes, the cabinet and published image endpoints support generation, while editing is available for models that publish that capability. Select a current model and options under Images API.
Is video generation available?
Yes, video generation is available in the cabinet and through asynchronous POST /v1/videos; GET /v1/videos/models returns current models. Choose a model from the public model catalog, create one job, poll its status, and download completed output under the cabinet video documentation.
Are public embeddings, audio, and web search available?
Public /v1 does not currently publish embeddings, audio transcription or translation, or Responses web_search endpoints. Web search is available as a cabinet chat capability, and external agents can use their own tools; see Coding tool integrations.
Teams and business billing
How do I create an organization and invite colleagues?
Create an organization in Settings, choose a role, and send an email invitation or one-use link. Workspace switching and roles are explained in Teams and organizations.
Can members see each other's chats?
Chats belong to the organization, and members with chat-read access can see its conversations; the Billing role does not receive that access. Use a personal workspace or a separate organization for isolation under Teams and organizations.
Are balance, keys, and usage shared in an organization?
They are scoped to the active organization: its keys spend its balance, and its requests appear in its Usage view. Use separate project keys for attribution, but not as a way to hide chats inside the same organization; see Teams and organizations.
How does an ООО or ИП receive an invoice?
Save an ООО or ИП payer profile in the active organization, then create and download the invoice from Balance. Statuses, legal links, and the business-document contact are in Business billing.
Support, account, and data requests
Where do I request technical support?
For account access, setup, and technical failures, use Telegram @provodai. Include the public code, model ID, time, client, and masked key prefix from API errors, never the complete key.
Where do I ask about company invoices and documents?
For companies, invoices, and closing documents, email info@provod.ai with the organization and invoice number. The Contact page lists the same channels without a response-time promise.
What if I cannot sign in?
Try the method already connected to the account, and use password recovery only for configured email sign-in. If that fails, send support the masked phone or email, sign-in method, time, and screenshot under Account and access.
How do I submit an account or personal-data request?
Email info@provod.ai and consult the current privacy policy. This documentation does not promise an identity-check method, deletion scope, or completion timeline.
Where are the legal documents and 152-FZ information?
Current pages include the offer, privacy policy, company details, and separate 152-FZ information. Link to those current pages instead of copying their text into a guide.