Sign in

OpenAI

Run GPT on Terminus

OpenAI's GPT models remain the most widely deployed general-purpose models in the world, and the OpenAI API shape is the lingua franca of the ecosystem — including Terminus's own gateway.

Terminus carries OpenAI directly and through Azure AI Foundry where that lane is configured. Use GPT in the desktop chat, behind standing agents, or from apps — and because the gateway speaks the OpenAI protocol, anything built against it can switch to any other provider by changing one string.

From the gateway

# The models your account can use
curl -s "$TERMINUS_API/v1/models?modality=chat" \
  -H "authorization: Bearer $TOKEN"

# Route a chat completion to one
curl -s "$TERMINUS_API/v1/terminus/chat/completions" \
  -H "content-type: application/json" \
  -H "authorization: Bearer $TOKEN" \
  -d '{
    "model": "@terminus/<address from the list above>",
    "messages": [{"role": "user", "content": "hello"}]
  }'

The gateway speaks the OpenAI chat-completions protocol.

One gateway, one bill

The Terminus gateway is OpenAI-compatible and meters every request. Switch providers by changing the model string — no new keys, SDKs, or invoices per vendor.

Agents, not just chat

Any model the gateway carries can power a Terminus agent: an autonomous worker with its own address, schedule or trigger, and scoped tool grants, running on Terminus infrastructure around the clock.

A desktop over your files

Conversations and agent runs happen in a workspace with a real file system — what the model produces lands in directories you browse, sync, and export, not in a chat log.

Questions

Which OpenAI models does Terminus carry?
The catalog tracks OpenAI's current models; /v1/models lists exactly what your account can use, each model by its address.
Is the Terminus gateway OpenAI-compatible?
Yes — /v1/terminus/chat/completions speaks the OpenAI chat-completions protocol with streaming and tool use, and /v1/terminus/embeddings answers an embedding model's vectors in OpenAI's Embeddings format, so OpenAI-shaped clients work against it directly.
Can I switch an agent from GPT to another model later?
Yes — the model is one field in the agent's configuration. Metering is per request, so you can compare what the same agent costs on different models.

Terminus is in beta. Join the waitlist, or write to early-access@terminus.build to become a founding user.