API overview

SayGM exposes several familiar HTTP interfaces through one gateway.

Interface	Base URL	Primary endpoint
OpenAI-compatible	https://api.saygm.com/v1	/chat/completions, /responses
Anthropic-compatible	https://api.saygm.com	/v1/messages
Gemini-compatible	https://api.saygm.com	/v1beta/models/{model}:generateContent
OpenAI Images	https://api.saygm.com/v1	/images/generations

Model IDs are provider-free. The gateway chooses an upstream route for each request; use the routing headers to prefer the cheapest or fastest route, or to limit which upstream providers may serve it.

Supported surfaces

• OpenAI Chat Completions, streaming and non-streaming.
• OpenAI Responses.
• Anthropic Messages, streaming and non-streaming.
• Gemini generateContent and streamGenerateContent with SSE.
• OpenAI Images generation at /v1/images/generations for models whose api_shapes include images.generations.
• OpenAI-compatible model listing and retrieval at /v1/models.
• Current SayGM balance at GET /v1/credits.

The generated API reference documents the OpenAPI-described request and response shapes. The compatibility guides explain SDK setup and surface-specific behavior.

Check credits

curl -fsS "https://api.saygm.com/v1/credits" \
  -H "Authorization: Bearer $GM_API_KEY"

This authenticated read returns the key’s current spendable balance. See check balance and current spend for the response fields. For history, see Usage.

Request lifecycle

Requests enter a gateway running inside an Intel TDX trusted execution environment. The gateway validates the SayGM key, checks available prepaid credit, selects an eligible route under the request’s routing mode, applies configured guardrails, forwards the request, and settles actual usage.

Versioning

The compatibility paths currently use the upstream interfaces’ versioned conventions. Avoid depending on undocumented fields. Additive SayGM metadata may appear on discovery responses.