Helicone Request Logger

Helicone is an open-source LLM observability platform. It sits as a proxy between your application and LLM providers (OpenAI, Anthropic, etc.), logging every request with latency, token usage, cost, and custom metadata — without changing your code beyond a URL swap.

Instead of calling api.openai.com, you call oai.helicone.ai. Add your Helicone API key as a header. Helicone forwards the request, logs everything, and returns the response. Latency overhead is <10ms. You can also use Helicone's SDKs for more advanced features like caching and rate limiting.

You can add any key-value metadata via Helicone-Property-* headers. Common uses: user ID, session ID, feature flag, A/B test variant, environment (prod/staging), model version. These appear in the Helicone dashboard for filtering, segmentation, and cost attribution per user/feature.

Yes — Helicone supports OpenAI, Anthropic, Google Gemini, Cohere, Together AI, Groq, and more. For Anthropic, use anthropic.helicone.ai instead of api.anthropic.com. For other providers, Helicone has gateway mode and custom mappings.

Helicone can cache LLM responses to reduce costs. Add 'Helicone-Cache-Enabled: true' header. Identical prompts return cached responses (saved to Cloudflare edge). You can also set cache TTL and seed for deterministic caching. Reduces costs significantly for repeated/similar queries.