POST /v1/messages.
Include these headers:
Parameters
| Parameter | Required | Meaning |
|---|---|---|
model | Yes | Exact Claude model ID available to your key. |
messages | Yes | Current input and conversation history. |
max_tokens | Yes | Maximum generated output length; must be at least 1. |
system | No | System prompt. |
thinking | No | Thinking configuration, only where supported by the selected model. |
messages request with incompatible fields unchanged to this endpoint.
Claude Code setup uses the origin https://infistar.ai, without /v1; the client adds the request path itself.