Skip to main content
calado watches your AI agent in production and tells you where its behavior is going wrong. Wrap your Anthropic or OpenAI client with two lines of code — or attach the Python callback handler to your LangChain or LangGraph agent — and calado captures every conversation in the background, then surfaces recurring issues you can act on.

Quickstart

Wire up the SDK, capture your first conversation, and run analysis.

Ingestion

Pick your path: Node.js SDK, LangChain handler, Direct API, or Serverless.

​
What you get

  • Automatic capture. Every call to messages.create or chat.completions.create is recorded. Nothing to annotate, no custom events to fire.
  • Analysis on your schedule. Conversations are stored on ingest; classification only runs when you click Run Analysis on the agent page, so you control when LLM cost is incurred.
  • Pattern-first. calado groups similar issues across conversations so you see what’s systematic, not per-conversation noise.
  • Out of the critical path. calado sits next to your agent. If our transport fails, your app doesn’t.

Get started

The fastest way in is the Node.js SDK.