AI Gateway

Every model behind one endpoint.

A unified gateway that routes chat, embeddings, and image generation across OpenAI, Anthropic, Google, and xAI through a single edge URL. Name the model, get the native response, and pay per call from your fluid wallet — observability built in.

model routing

Pick the model, skip the plumbing

The gateway maps your model name to the right upstream, signs the request with a credential that never leaves the platform, and streams the answer back in the provider's own format. Swapping from one model to another is a one-line change.

  • One URL fronting OpenAI, Anthropic, Google, and xAI
  • Streaming and native response shapes preserved

one plane, many providers

A single plane across vendors

Chat, embeddings, and image generation all enter through the same plane, so your tier policy, rate limits, and wallet rules apply uniformly. Add a provider on the platform side and it shows up at the same endpoint — no new SDK, no new key.

  • Uniform auth, gating, and limits per call
  • New upstreams appear without client changes

pricing

Pay for what each call costs

Model calls are fluid-only: metered per call and charged to your shared wallet the moment the request resolves. Tiers and grants live on the pricing page. Two numbers tell the story:

providers

4 behind one endpoint

failed calls

$0.00 / never billed

Point your client at a single gateway URL and name the model with a company-prefixed alias — openai/gpt-5-nano, anthropic/claude-haiku-4.5, google/gemini-3.1-flash-lite, xai/grok-4.3. The gateway resolves the provider, attaches the managed upstream credential, and returns the native response shape. Switching models is a string change, not an integration.

Related

Unified endpoint, Model aliases, Native responses, OpenAI · Anthropic · Google · xAI

For more information, ask your agent.