Connect Claude Code to Anoman AI

Step 1: Configure Claude Code CLI

Set two environment variables before launching Claude Code. The CLI reads these automatically.

export ANTHROPIC_BASE_URL=https://api.anoman.io/anthropic
export ANTHROPIC_API_KEY=anm-sk-your-key-here

# Launch Claude Code — all requests now route through Anoman
claude

Step 2: Python SDK configuration

Pass base_url when constructing the Anthropic client. The SDK sends all requests to the Anoman proxy, which applies guardrails before forwarding to Anthropic.

import anthropic

client = anthropic.Anthropic(
    api_key="anm-sk-your-key-here",
    base_url="https://api.anoman.io/anthropic",
)

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Summarize this document."}],
)
print(message.content[0].text)

Streaming responses

Streaming works identically through the Anoman proxy. Guardrails run on the prompt before the stream opens.

with client.messages.stream(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain this code."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

When a guardrail triggers

If a request is blocked by the guardrail pipeline, Anoman returns a 403 before any content reaches Anthropic. The SDK surfaces this as an anthropic.APIStatusError.

try:
    message = client.messages.create(...)
except anthropic.APIStatusError as e:
    if e.status_code == 403:
        print("Blocked by guardrail:", e.body)
    raise

Check docs/openapi/ for the current error schema.

Guardrail transparency headers

Every response from the Anoman proxy includes guardrail result headers. Inspect these in your logs to verify the pipeline is active.

X-Anoman-Guardrail-Injection: pass score=0.04
X-Anoman-Guardrail-Pii: pass
X-Anoman-Guardrail-Content: pass
X-Anoman-Cache: none