Claude Code → OrcaRouter Lite

Lite serves the Anthropic Messages API natively (POST /v1/messages), so Claude Code and the official anthropic SDKs connect without changes.

Claude Code

export ANTHROPIC_BASE_URL=http://localhost:8000    # no /v1 suffix
export ANTHROPIC_API_KEY=sk-orca-...               # your Lite key
# optional: pin the model (any catalog model, or "auto")
# export ANTHROPIC_MODEL=claude-3-5-sonnet-latest

claude

Requests route through Lite's normal pipeline: model="auto" works, the prompt cache works, and every request lands in the local dashboard.

Anthropic Python SDK

from anthropic import Anthropic

client = Anthropic(
    base_url="http://localhost:8000",   # SDK appends /v1/messages itself
    api_key="sk-orca-...",
)
message = client.messages.create(
    model="auto",                        # or any catalog model
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)
print(message.content[0].text)

Streaming (client.messages.stream(...)) and tool use work as usual.

Known limitations (v1)

• thinking (extended thinking) is accepted but dropped — the request still works, without a thinking budget. A structlog warning is emitted.
• top_k is dropped; temperature / top_p pass through, and values outside the API's own 0–1 range are rejected with a 400 rather than forwarded (an upstream rejection would reach you as a retryable 500). max_tokens must be >= 1, likewise.
• stop_sequences beyond the OpenAI wire cap of 4 are truncated to the first 4 (a structlog warning is emitted).
• Images inside a tool_result are forwarded as a user message following the tool result, since the internal format's tool messages are text-only. Non-text blocks in system (which the API itself does not allow) are dropped with a warning.
• Streaming message_start.usage.input_tokens is the same estimate /v1/messages/count_tokens returns — the upstream only reports the exact count at end-of-stream, where it is sent in message_delta.
• Server-side tools (web search, computer use), PDF document blocks, and the Files API return 400. Custom function tools are fully supported.
• POST /v1/messages/count_tokens returns an estimate (litellm.token_counter, chars/4 fallback), good enough for Claude Code's context tracking.