Anthropic

Claude API Audit Logging & Evidence

Direct Claude API evidence at the Messages boundary

Anthropic Messages API integration mirrors the OpenAI wrapper pattern: capture inference boundaries, tool use, and commitments for GDPR-aligned EU deployments.

Where Sigigo sits

At the Anthropic Messages API boundary — before requests leave your VPC and after responses return. The wrapper maps Claude-specific metadata (model id, stop reason, tool blocks) into the same ai_act schema used across providers.

Common deployment paths

Claude powers a large share of EU agent and copilot stacks — from direct API integrations to Bedrock in eu-central-1:

  • Direct Messages API for apps, agents, and internal tools
  • Legal, finance, and enterprise copilots with high-risk Article 12 logging when Annex III applies
  • Multi-turn threads with tool use and streaming
  • Composable with LangChain and OpenRouter without duplicate emits

Integration shape

@sigigo/providers-anthropic — same core proof library as OpenAI, thin adapter for Messages + tool streaming.

  • wrapAnthropic(client, sigigo) — drop-in client proxy
  • Shared session_id across multi-turn threads
  • Commitment processor for system prompts and retrieved context

Evidence events

Canonical event types emitted through this integration path:

  • ai_act.inference_started / inference_completed
  • ai_act.input_received (commitment)
  • ai_act.output_generated
  • ai_act.action_executed

Related

Integration availability may vary by tenant, region, and deployment model. This page describes product architecture — not a guarantee of feature availability. See compliance disclaimers and contact us for early access.