Messages
The Anthropic Messages format, at POST /v1/messages.
Install the SDK with pip install anthropic or npm install @anthropic-ai/sdk. max_tokens is required. The system prompt is the top-level system parameter; messages takes no system role.
curl https://raon-api.krafton.ai/v1/messages \
-H "x-api-key: $RAON_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "raon-1-mini-preview",
"max_tokens": 1024,
"system": "You are a helpful assistant.",
"messages": [{"role": "user", "content": "Hello!"}]
}'The answer starts with a thinking block, then the text block. stop_reason is end_turn when the answer is complete and max_tokens when it reached the limit.
The Anthropic SDKs add /v1 themselves, so their base URL has none. The same goes for ANTHROPIC_BASE_URL in Claude Code.
Tool selection
tool_choice takes auto, any or none. Naming one tool (tool_choice.type: "tool") is refused with 400; omit tool_choice or use any instead.