Back to guidesGetting started
Instrumenting your first LLM call
Install the SDK and get token, cost, and latency data flowing from your first agent or chat endpoint in under ten minutes.
10 min to complete
1
Install the SDK
Run `npm i @trasys/sdk` in your service and set the `TRASYS_API_KEY` environment variable from your project's API Keys page.
2
Wrap your LLM client
Pass your existing OpenAI, Anthropic, or Groq client into `trasys.wrap()` — no changes to how you call the model are required.
3
Tag the call
Add an optional `agentName` and `sessionId` to each call so it shows up correctly in AI Monitoring's agent breakdown.
4
Confirm data is flowing
Open the AI Monitoring overview for your project — the first call typically appears within 30 seconds.