Helicone
Free tierAI Gateway & LLM Observability for building reliable AI applications
Free tier available·Technical·API available·Open source
Key strengths
Unified AI gateway supporting OpenAI, Anthropic, Azure, and many more providersLLM observability with request logging, session tracking, and user analyticsPrompt management with datasets and playground for iterative improvementRate limiting and alerting for production reliabilityFast integration — route traffic through Helicone with a single line change
Free tier + paid plans
San Francisco, United States
Founded 2023
Self-hostable
No ratings yet
Technical Integration
Base URL Proxy (OpenAI Example)
The simplest integration is a single-line base URL change:
import openai
client = openai.OpenAI(
api_key="YOUR_OPENAI_API_KEY",
base_url="https://oai.helicone.ai/v1",
default_headers={
"Helicone-Auth": "Bearer YOUR_HELICONE_API_KEY"
}
)
Supported Providers
- OpenAI —
oai.helicone.ai - Anthropic —
anthropic.helicone.ai - Azure OpenAI —
openai.azure.comvia Helicone proxy - LiteLLM, Anyscale, Together AI, OpenRouter — all supported
Key Features for Developers
- Request logging — Full prompt/completion capture with metadata tagging (
Helicone-User-Id,Helicone-Session-Id, custom properties) - Segments & HQL — Filter requests by user, model, cost, latency, or custom properties using Helicone Query Language
- Datasets & Playground — Curate request datasets and run prompt experiments directly from the UI
- Rate Limiting — Set per-user or per-key rate limits via request headers
- Alerts — Webhook or email alerts on cost thresholds, error spikes, or latency anomalies
- Self-hosting — Open-source repo available for teams that require on-premises deployment