Choose a connection method
Obsivara receives events over HTTP. Pick the method that fits your stack — no install required for webhook and OTLP.
Method comparison
| Your situation | Recommended method | Page |
|---|---|---|
| I can send an HTTP POST from any language or tool | Generic webhook | Generic webhook → |
| I already use OpenTelemetry in my stack | OTLP traces endpoint | OpenTelemetry (OTLP) → |
| I want an AI agent to wire up telemetry for me | Agent runbook | Connect your AI agent → |
| I use n8n | n8n webhook node | n8n → |
| I use LangSmith | LangSmith webhook | LangSmith → |
| I use Langfuse | Langfuse webhook | Langfuse → |
| I write TypeScript and want typed helpers | JavaScript SDK (preview) | JavaScript SDK → |
Start with the webhook
The generic webhook is the simplest path: one POST to https://api.obsivara.com/ingest/webhook/generic with a JSON body records a run event. No SDK, no library, no dependency — just HTTP. It works from curl, Python, Node, Go, or any workflow tool.
If your stack already emits OpenTelemetry traces, point your exporter at the OTLP endpoint instead — Obsivara maps gen_ai.* attributes automatically, so you get LLM cost and token tracking with no extra work.
Let an AI agent do it
The agent runbook is designed to be given directly to an AI agent (e.g. Claude, GPT-4o). It contains the exact endpoints, payloads, and verification steps the agent needs to instrument your workflow end-to-end. Copy the page URL or the machine-readable version into your agent's context.
First — get an API key
Every method requires an obs_live_... API key with the ingest:write scope. See Authentication for how to create one. The key goes in the Authorization: Bearer obs_live_... header on every request.