Skip to content

Models API

Annota AI offers the chat models of the Models catalog through the Models API, compatible with OpenAI Chat Completions. Any OpenAI client or SDK works: set the base URL and an API key of your organization.

https://api.annota.ai/openai/v1

The API supports:

  • GET /models: lists the catalog models. It needs no API key.
  • POST /chat/completions: sends a chat request, with or without streaming. It needs an API key.

Use the model ID shown in the catalog, such as deepseek-v4.1-flash-1017.

API keys belong to the organization, not to a member, and spend its credits until they are revoked. Only owners and admins can manage them. Other members can ask an organization owner or admin for a key.

  1. Go to Settings > API keys and select Create API key.
  2. Enter a Name that tells where the key is used, such as a project or a server, and select Create API key.
  3. In Copy your API key, copy the key and store it somewhere safe, then select I saved the key.

The API keys page lists each key with its name, the masked key, who created it, when, and when it was last used. An organization can hold a limited number of keys: revoke one before creating another.

The examples read the key from the ANNOTA_AI_API_KEY environment variable. Each model page in the app has the same examples in cURL, Python, TypeScript, Go, and Rust under API.

Terminal window
curl https://api.annota.ai/openai/v1/chat/completions \
-H "Authorization: Bearer $ANNOTA_AI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4.1-flash-1017",
"messages": [{"role": "user", "content": "Hello"}]
}'

Each request is priced exactly at the rates of the model it uses, per million input, cached input, and output tokens, and paid with organization credits. Requests never count toward a member’s weekly Brain usage limit. A request needs a positive credit balance.

  • Usage appears as OpenAI-compatible API in the billing usage. See Credit usage.
  • If a client disconnects during a streamed response, what the model already generated is still charged.

See Models for the rates.

  1. Go to Settings > API keys.
  2. Select Revoke on the key’s row and confirm.

Requests with the key stop working at once. This cannot be undone. Revoking a key is the way to replace it if it may have leaked.