What it does
This is the code repository for Helicone, an AI gateway and LLM observability platform. The gateway reaches more than 100 models through an OpenAI-style API with one key, with routing and automatic fallbacks. Logging works by changing the base URL in your existing code, and it supports providers and frameworks such as OpenAI, Anthropic, LangChain, Gemini and the Vercel AI SDK. You can inspect traces and sessions for agents, chatbots and document pipelines, and track cost, latency and quality. Data can be exported to PostHog for custom dashboards. A playground lets you test prompts, and prompt management versions prompts using production data and deploys them through the gateway without code changes. Fine-tuning is offered through partners OpenPipe and Autonomi. You can self-host with Docker using the included compose file. A Helm chart for enterprise workloads is available by contacting the team. The project is licensed under Apache 2.0, and the repository lists SOC 2 and GDPR compliance. The hosted service has a monthly free tier of 10,000 requests with no credit card, and you add credits for more usage.
Capabilities
- AI gateway to 100+ models through an OpenAI-style API with fallbacks
- One-line logging integration for major providers and frameworks
- Traces, sessions and cost, latency and quality metrics
- Playground and prompt management deployed through the gateway
- Self-hosting with Docker, plus a Helm chart for enterprise workloads
- Apache 2.0 licence
When to use it
- You want to log and debug LLM requests without large code changes
- You want to run your observability stack yourself rather than use a hosted service
- You need one API and fallbacks across many model providers
