[01] Products

Pick your control plane

One platform, two deployment models. Both serve the same OpenAI-compatible API from your own hardware — the difference is who runs the control plane.

[02]Deployment models

[A]

OPEN →

Cloud

Managed · pay-as-you-go

The full managed platform. Auto-scaling API instances, Redis-routed conduit connections, credit billing and the complete console. Connect hardware from anywhere and serve a public or private endpoint.

[B]

OPEN →

Self-hosted

Single service · annual license

Console and API combined into one Docker service on your infrastructure. No billing, no metering, no dependency on our platform — choose a license tier and own the whole loop.

[03]Shared stack

Conduit

Self-hostable agent connecting vLLM, llama.cpp and other engine servers on your machines.

Endpoints

Public or private inferencing endpoints with composable routing to your sources.

Models

Serve any local model; curated recommended-models catalog included.

API

OpenAI-compatible REST surface — drop-in for existing agents and tooling.

Tool services

Server-side tool interception and MCP tool services wired into requests.

Console

Sources, models, endpoints and spend managed from one control panel.