FAQ

Common questions about Kalibr.

General

What is Kalibr?

Observability for multi-agent AI systems. Track costs, debug failures, and monitor performance across OpenAI, Anthropic, and Google.

How is it different from LangSmith?

LangSmith is for prompt engineering with LangChain. Kalibr is for production cost tracking with any framework.

What frameworks does it work with?

Any code using OpenAI, Anthropic, or Google SDKs. This includes LangChain, LlamaIndex, CrewAI, and custom code.

Setup

How long does setup take?

5 minutes. Install, set API key, import.

What code changes are needed?

One line: import kalibr before your LLM imports.

Does it affect performance?

<1ms overhead per call. Traces are batched and sent async.

Pricing

What counts as a trace?

One workflow execution. A workflow with 5 LLM calls = 1 trace.

Troubleshooting

Traces not appearing

1. Check API key: echo $KALIBR_API_KEY
2. Check collector URL: echo $KALIBR_COLLECTOR_URL (should be https://api.kalibr.systems/api/ingest)
3. Check import order: import kalibr first
4. Check connectivity to api.kalibr.systems

Support

• Dashboard: dashboard.kalibr.systems
• Docs: docs.kalibr.systems
• Email: support@kalibr.systems