#API reference

Geekit exposes each upstream provider's own request format and replaces only the authentication with a Geekit-issued token. Point your SDK's base URL at the gateway, swap the key for a Geekit token, and the rest of your code is unchanged.

#Authentication

Authorization: Bearer sk-xxxxxxxxxxxxxxxx

Tokens are created in the dashboard and can be restricted by group, allowed models, quota, and expiry.

#OpenAPI specifications

The machine-readable specifications are published alongside these docs and can be imported into Postman, Insomnia, or a code generator:

#Supported request formats

#Format conversion

The gateway converts between formats, so a client speaking one format can reach an upstream that speaks another:

• OpenAI Compatible ⇄ Claude Messages
• OpenAI Compatible → Google Gemini
• Google Gemini → OpenAI Compatible (text only; function calling is not supported yet)
• Reasoning content folded into regular content

#Example

curl https://<your-gateway-domain>/v1/chat/completions \
  -H "Authorization: Bearer $GEEKIT_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

#Billing

Charges follow the usage reported by the upstream, separating input, output, cache-hit, and reasoning tokens, then converting to quota with the channel's ratios. Every request writes its billing detail to the log, so you can reconcile by token, model, or channel in the dashboard.