Anthropic

Use this provider to route to Anthropic's Claude models. Aurora supports both the OpenAI-compatible translation path (/v1/chat/completions → Claude) and the native Anthropic ingress (/v1/messages for the Anthropic SDK).

Activate

ANTHROPIC_API_KEY=sk-ant-...

Restart Aurora. The anthropic provider appears at /v1/models with the current Claude catalog.

To use the native Anthropic SDK against Aurora (recommended for Claude-only apps), also enable the ingress:

ENABLE_ANTHROPIC_INGRESS=true

This exposes /v1/messages and /v1/messages/count_tokens. See Anthropic Messages API.

Default models

Override the catalog with ANTHROPIC_MODELS=claude-sonnet-4-5,claude-haiku-4-5.

Multiple instances

ANTHROPIC_PROD_API_KEY=sk-ant-...
ANTHROPIC_STAGING_API_KEY=sk-ant-...

Aurora exposes them as anthropic-prod and anthropic-staging.

Custom base URL

ANTHROPIC_BASE_URL=https://anthropic-proxy.example.com/v1

Use this for AWS Bedrock-hosted Claude, GCP Vertex-hosted Claude, or a corporate proxy.

Gotchas

• Prompt caching — Claude's cache_control blocks are honored end-to-end. Aurora forwards them and reports cache hit/miss in the response usage.
• Reasoning / extended thinking — Claude's thinking content blocks are preserved. Aurora exposes them as a separate field in the OpenAI-shaped response.
• Tool use — full support, including tool-result messages, parallel tool calls, and tool choice. No provider-specific translation needed.
• Max tokens — Claude requires max_tokens and will error without it. Aurora injects a default if you forget.
• System prompt — Claude accepts a top-level system field. Aurora's guardrail system_prompt_inject / system_prompt_override features target this exact field. See Guardrails.

Sample request (OpenAI format)

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AURORA_MASTER_KEY" \
  -d '{
    "model": "anthropic/claude-sonnet-4-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello from Aurora"}]
  }'

Sample request (native Anthropic SDK)

from anthropic import Anthropic
 
client = Anthropic(
    base_url="http://localhost:8080",
    api_key="sk-aurora-...",
)
msg = client.messages.create(
    model="anthropic/claude-sonnet-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello from Aurora"}],
)
print(msg.content[0].text)

Related

• Anthropic Messages API — /v1/messages and /v1/messages/count_tokens
• Anthropic SDK Integration — point the official SDK at Aurora
• Providers Overview