Anthropic SDK
Use the official Anthropic SDK for Python or Node with Nexrouting through the Messages API at /v1/messages.
La documentazione è in inglese.
POST /v1/messages speaks the Anthropic Messages format. Point the Anthropic SDK at https://api.nexrouting.com, without /v1: the SDK adds /v1/messages itself. The SDK sends your key as x-api-key, which the gateway accepts.
Install
pip install anthropicSend a message
import osfrom anthropic import Anthropic client = Anthropic( base_url="https://api.nexrouting.com", api_key=os.environ["NEXROUTING_API_KEY"],) message = client.messages.create( model="anthropic/claude-sonnet-5.5", max_tokens=1024, messages=[{"role": "user", "content": "Hello"}],)print(message.content[0].text)Or set ANTHROPIC_BASE_URL and ANTHROPIC_API_KEY in the environment; both SDKs read them.
Which models
Only models that list /v1/messages in their endpoints accept this format: the Anthropic models and a number of others. Check GET /v1/models. Any other model returns 400 unsupported_endpoint; use it through /v1/chat/completions instead.
Version and beta headers
The SDK sets anthropic-version; the gateway passes it on to the model provider, and uses 2023-06-01 when it is missing. anthropic-beta is passed on as well. Whether a beta feature works depends on the model.
Streaming
with client.messages.stream( model="anthropic/claude-sonnet-5.5", max_tokens=1024, messages=[{"role": "user", "content": "Write a haiku about routing."}],) as stream: for text in stream.text_stream: print(text, end="")Prompt caching
cache_control on your content blocks is passed to the model provider. Writing to the cache is billed at a multiple of the input rate (1.25× for 5 minutes, 2× for 1 hour). Reading from the cache is billed at the normal input rate for now. See Pricing & usage.
Errors
Errors on /v1/messages use the Anthropic shape ({"type": "error", "error": {...}}), so the SDK raises its usual exceptions, such as RateLimitError for 429. See Errors.
Not available
Token counting (/v1/messages/count_tokens) and Message Batches return 404 not_found. GET /v1/models returns the OpenAI-style list described in List models, which the Anthropic SDK's models.list() may not read correctly; call it directly instead.