Helicone

Helicone provides observability and analytics for large language model (LLM) usage, capturing telemetry, token usage, latency, and request metadata through a proxy or SDK and exposing results in a web dashboard and API. It helps teams monitor LLM performance, troubleshoot failures, and understand cost-driving usage patterns. Helicone is designed to be deployed self-hosted or used as a hosted service to collect LLM metrics and logs.

• Telemetry capture — Use Helicone as a proxy or instrument the SDK to capture requests, responses, tokens, and latency for LLM calls.
• Dashboard and API analytics — View aggregated metrics, token consumption, and latency trends in the web dashboard or query metrics programmatically via the API.
• Request logging and replay — Store request/response pairs and metadata for debugging and post-hoc analysis.
• Integrations and SDKs — Install language- or framework-specific SDKs or route traffic through the Helicone proxy to start collecting data.
• Self-host or hosted — Deploy the open-source components to your infrastructure for local data control or use the hosted offering for managed telemetry.

To get started, create an account or deploy the open-source components, configure your LLM client to use Helicone's proxy or SDK, and open the dashboard to inspect metrics and logs. Use the API for automated metric queries and export.