Inspect prompts, trace multi-turn conversations, and catch issues before they hit production. Full request/response logging with zero code changes. Need observability without the proxy hop? Trace mode sends requests directly to your provider and streams metadata async.
Grepture logs every LLM request and response flowing through your app. You get a real-time traffic log showing prompts, completions, token counts, latency, and status codes — across every provider and model. In trace mode, requests bypass the proxy entirely and go straight to the provider. The SDK captures metadata (tokens, model, latency) and sends it async — zero latency overhead, full dashboard visibility.
Point your SDK at Grepture (one config line). Every request is captured, indexed, and searchable in the dashboard. Filter by model, endpoint, status, or time range. Trace full conversation threads across multiple requests. For latency-sensitive workloads, set mode: "trace" — requests go directly to the provider while trace data flows to your dashboard in the background.
| Time | Method | URL | Status | Cost | Duration |
|---|---|---|---|---|---|
| 2s ago | POST | /v1/chat/completions | 200 | $0.0342 | 1.2s |
| 5s ago | POST | /v1/embeddings | 200 | $0.0001 | 84ms |
| 12s ago | POST | /v1/chat/completions | 200 | $0.0189 | 2.4s |
| 18s ago | POST | /v1/messages | 429 | $0.0000 | 12ms |
| 31s ago | POST | /v1/chat/completions | 200 | $0.0024 | 890ms |
| 45s ago | POST | /v1/chat/completions | 200 | $0.0510 | 3.1s |
npm install @grepture/sdk — wrap your client — done.
Free for up to 1,000 requests/month · No credit card required
Get Started Free