Helicone

An AI gateway and LLM observability platform with self-hosted Apache 2.0 monitoring, Ollama integration, agent tracing, and cost tracking.

Helicone helps developers inspect and debug LLM applications, including agents, chatbots, and document processing pipelines. It combines request monitoring with a gateway for routing calls across AI providers.

Request and agent tracing: Inspect logged requests, traces, and sessions to investigate errors across a sequence of model calls.
Provider routing: The hosted gateway provides an OpenAI-compatible API with routing and automatic fallbacks for external providers, including OpenAI, Anthropic, and Gemini.
Cost and performance analysis: Track spending, latency, and quality, or export analytics to PostHog for custom dashboards.
Prompt testing and management: Use the playground to test prompts against sessions and traces, keep prompt versions informed by production data, and deploy prompts through the gateway without changing application code.

The open source observability platform can run on your own server under Apache 2.0; Helicone also offers a hosted service. Its Ollama integration supports applications that run models locally. Calls through the hosted gateway go to external AI providers, so self-hosting observability doesn't make those model requests local.

Helicone connects with LangChain, LlamaIndex, LangGraph, and the Vercel AI SDK. RAGAS supports evaluation workflows, while OpenPipe and Autonomi provide partner integrations for fine-tuning.

LLM Observability and Tracing Prompt Management and Playgrounds Web LLM tracing Ollama integration Prompt versioning