Quickstart
From zero to your first response in three steps: set the base URL, create an API key, then send a request.
Base URL
OpenAI-format clients use a base URL that ends with /v1. Anthropic-format clients use the root without /v1, because their SDK appends /v1/messages itself.
| Setting | Value | Used by |
|---|---|---|
| base_url | https://ai.geraikita.com/v1 | OpenAI SDK, Cline/Roo (OpenAI Compatible), OpenCode, n8n, Continue |
| base_url | https://ai.geraikita.com | Anthropic SDK, Claude Code CLI, Cline (Anthropic) |
| api_key | gk-… | All clients — keys are created in the Dashboard |
Create an API key
- Sign in to Geraikita AI.
- Add GeraiPoints on the Top up page (or claim a free key from Bansos).
- Open the Dashboard → Create API key. Copy the
gk-…key — the raw key is shown only once.
export GERAIKITA_API_KEY=gk-…, instead of hard-coding it.First request
Pick the language or SDK you use. All examples call the claude-sonnet-5 model; replace it with another model id from the model list.
curl https://ai.geraikita.com/v1/chat/completions \
-H "Authorization: Bearer gk-xxxxxxxx" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5",
"messages": [{"role": "user", "content": "Hello!"}]
}'The response is JSON in the same shape as the OpenAI API (or Anthropic, when using /v1/messages), including a usage object with the number of billed tokens.
Stream the response
For long responses and agentic tools, enable stream so chunks of the response flow in immediately. The server sends headers right away and keepalive signals while waiting for the first token. Set your app’s timeout to at least 300 seconds.
curl https://ai.geraikita.com/v1/chat/completions -N \
-H "Authorization: Bearer gk-xxxxxxxx" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5",
"stream": true,
"messages": [{"role": "user", "content": "Hello!"}]
}'Next steps
- Set it up in Claude Code, Cline, OpenCode, and other tools: Tool integrations.
- Using an agentic tool? Read Agentic tool settings before sending large prompts.
- See the available model ids: Models.
- Handle errors correctly: Limits & error codes.