Documentation
GateML is an OpenAI-compatible API gateway. Point your existing code at https://api.gateml.io/v1 and your requests are routed, observed, and protected — with automatic fallback on failures.
New here? Start with the Quickstart →
What GateML does
When your code calls POST /v1/chat/completions with a GateML key, we:
- Validate your API key (test or live)
- Route to your configured primary model (e.g. gpt-4o)
- On 429 or 5xx, retry the next model in your fallback chain
- Log the request — latency, tokens, cost, status
- Return the response in standard OpenAI format
Key concepts
Test keys (gml-sk-test_...)
Return synthetic responses. No real LLM calls. No charges. Use during development.
Live keys (gml-sk-live_...)
Forward to real providers using your stored provider API keys.
Provider keys
Your OpenAI / Anthropic / Google API keys, stored encrypted (AES-256) in GateML.
Fallback chain
Ordered list of models to try on failure. Configure in Gateway → Routing.
Quicklinks