One connection.
More possibilities.

Explore the API Talk to us

Model choice, thoughtful routing, and usage you can see.
A gateway for the applications you build.

In development · Public API access is not available yet

210+

Provider-neutral by design. Launch catalog to be announced.

Built for the way you build.

The gateway we’re working toward.

#1 Connect Planned

One API. More models.

A consistent chat-completions interface for model access. Start with text and streaming, without separate integration logic for every provider.

#2 Route Planned

Choice with a reason.

Task-aware routing designed around capability, capacity and request constraints. Select an eligible model and inspect why it was chosen.

#3 Stream Planned

Keep the response moving.

Incremental responses over server-sent events. Bring streaming into your application through the same proposed API contract.

#4 Control Planned

Pin it. Or let it route.

Choose a model explicitly or allow automatic selection. A pinned model stays pinned unless you enable fallback.

#5 Inspect Planned

See every request.

Selected model, request usage and routing metadata in one place. Understand allowances and known capacity limits as you build.

#6 Build Planned

Your application. Your tools.

Use the gateway for your own applications. Agent is the first intended client; tools and project memory remain under the client’s control.

For developers

A familiar interface.
A new way to connect.

The proposed contract starts with a documented subset of chat completions: text, streaming, and explicit model selection.

See the launch plan

from openai import OpenAI

client = OpenAI(
    base_url="https://YOUR_CALYTHE_ENDPOINT/v1",
    api_key="YOUR_CALYTHE_KEY",
)

response = client.chat.completions.create(
    model="MODEL_ID",
    messages=[{"role": "user", "content": "Hello."}],
    stream=True,
)

Proposed API · Placeholder endpoint · No live requests

Opening in stages.

No launch date announced.

01 / Foundations

Verify the routes.

Provider admission, available capacity, and the initial API contract.

02 / Agent pilot

Connect the first client.

Validate adapters, routing, shared quota control, and request visibility.

03 / Developer access

Open the gateway.

Publish documentation, a verified catalog, and access limits after validation.

Developer questions

Good questions.
Clear answers.

Can I get an API key today?

Router is in development. Public API keys and a live endpoint are not available yet.

Which models will be available?

The launch catalog is not confirmed. Providers and endpoints will be verified before they are offered.

How will pricing work?

The initial plan is a capped free-only pilot. Paid pricing and broader access have not been announced.

Does Router run my agent’s tools?

The gateway handles model inference. Tools and project memory remain in your client, under that client’s permissions.

Can I use it without Calythe Agent?

Yes, that is the intended developer product. Agent is the first planned client, rather than a requirement for later access.

Build with Calythe.

Talk about Router Explore Agent

One connection.More possibilities.