Developers

Keys and guides for the OpenAI- and Anthropic-compatible API. Requests are paid from the same balance as chat.

Base URLs

Give OpenAI SDKs and OpenAI-compatible tools the first URL. Anthropic SDKs add /v1/messages themselves, so they take the shorter one.

OpenAI-compatibleOpenAI SDKs and most tools: https://hide.st/api/v1
Anthropic-compatibleAnthropic SDKs and Claude Code: https://hide.st/api

Authentication

Send your key in either header. Requests without a valid key get a 401.

Authorization: Bearer sk-hide-...

Keys are shown once, when you create them, and stored only as a hash. If you lose one, revoke it on the API keys tab and create a new one.

Endpoints

Paths are relative to https://hide.st. Request and response bodies follow the OpenAI or Anthropic format of the same endpoint.

Models

GET /api/v1/models: Text models in OpenAI's list format. Add ?detailed=true for each model's name, the prices you pay per 1M tokens, context_length, max_output_tokens and capabilities.
GET /api/v1/models/{id}: One model with full details. IDs contain a slash, as in /api/v1/models/anthropic/claude-fable-5.1.
GET /api/v1/image-models: Image generation models and their prices.
GET /api/v1/video-models: Video generation models and their prices.
GET /api/v1/3d-models: 3D generation models and mesh tools, and their prices.
GET /api/v1/audio-models: Speech and transcription models and their prices.
GET /api/v1/embedding-models: Embedding models and their prices.

Text

POST /api/v1/chat/completions: OpenAI Chat Completions, with streaming and tools. stream_options.include_usage is supported, and usage.cost is the charge in USD.
POST /api/v1/responses: OpenAI Responses, with function tools only. store is always false, so previous_response_id and background aren't available: send the full input each time.
POST /api/v1/messages: Anthropic Messages, with streaming, tool use, extended thinking and prompt caching.
POST /api/v1/messages/count_tokens: Counts the input tokens of a Messages request without running it.
POST /api/v1/embeddings: OpenAI Embeddings.

Images, audio and video

POST /api/v1/images/generations: OpenAI image generation.
POST /api/v1/audio/speech: Text to speech, in OpenAI's format.
POST /api/v1/audio/transcriptions: Speech to text, in OpenAI's format.
POST /api/generate-video: Starts a video or 3D job and returns its requestId. Send input files as base64 data URLs in imageDataUrl, lastImageDataUrl, videoDataUrl, audioDataUrl or, for 3D mesh tools, model_url; images takes more views for multi-view 3D models. Links to other sites aren't accepted. Files are cleaned and deleted when the job ends. Jobs take a while, so this call doesn't wait for the result.
GET /api/video/status?requestId=…: Status of a video or 3D job. Poll it until it's ready: videos return video_url, 3D models a GLB in model_url and every format in files. The links work for 24 hours, then the files are deleted.
DELETE /api/video/status?requestId=…: Deletes a finished job's files right away, once you've downloaded them.

Account

GET /api/v1/balance: Your balance and available balance in USD. Available is the balance minus holds for requests still running.

Model IDs

Model IDs have the form vendor/model, for example anthropic/claude-fable-5.1, openai/gpt-6-sol or meta-llama/llama-3.1-8b-instruct.

A short name works when it's unique, such as gpt-6-sol. Claude Code style names such as claude-fable-5-1 map to the matching Anthropic model. Browse every model and its price on the Models page.

Billing and request IDs

Each request first places a temporary hold on your balance. When it finishes, you're charged the model's price for what it used and the hold is released.

For some streamed responses the charge starts as an estimate. It's corrected automatically to the provider's exact cost, usually within minutes and always within 24 hours.

Responses from model endpoints carry an x-request-id header that starts with req_. It matches the request on the Usage page, so include it when you contact support about a request.

Not supported

These features need tools or stored state on the provider's side. Requests that use them fail with a 400 error that names the feature.

Web search: the :online model suffix and the webSearch and web_search_options parameters.
Server-side memory, including the :memory suffix.
Server tools in the Messages API: web_search, code_execution and the MCP connector. Tools that run in your own code work.
Hosted tools in the Responses API. Function tools work.
Stored responses: previous_response_id, background mode and conversations.