What you get
Single endpoint
One URL, one request format, and one response shape for every provider and model. Switch from
openai/gpt-4o to anthropic/claude-3-5-sonnet-20241022 by changing a single string.Three auth modes
Issue short-lived HMAC tokens from your backend (
GatelitSigned), pass OIDC tokens from Clerk, Supabase Auth, or Auth0 (Bearer), or use long-lived service keys for backend-to-backend calls (GatelitKey).Structured output
Request JSON responses from any supported model using
json_object or json_schema mode. Gatelit applies the correct provider-specific adapter automatically.Request logging
Every request is logged with token counts, estimated cost, latency, model used, and the auth identity — all visible in the dashboard without any instrumentation in your app.
Bring your own keys
Each organization can supply its own OpenAI, Anthropic, Google, and Mistral API keys. The gateway uses your org’s key for every request, keeping spend and quotas fully scoped to your account.
Saved prompts
Store versioned prompt templates in the dashboard and execute them by reference. Variables are resolved at request time, and every execution is correlated to the prompt in the logs.
Supported providers
Gatelit currently supports four LLM providers. Use any model from these providers by prefixing the model name with the provider identifier:Auth modes at a glance
Where to go next
Quickstart
Make your first authenticated LLM request in under five minutes.
How it works
Understand the full request lifecycle — auth, provider routing, BYOK, and logging.
Authentication
Choose the right auth mode and learn how to issue tokens.
SDK reference
Full reference for
GatelitClient, signToken, and the Conversation helper.