Engine for Weights, AI Compute & Serving

EWACS

Run open models on your own machine.

A desktop app that speaks the Ollama and OpenAI APIs unchanged — with a commercial-use-aware marketplace behind it, so every model you pull is licensed to ship. Install it, sign in, and chat.

Engine
Weights
AI
Compute
Serving
DROP-IN COMPATIBLE
POST /api/generate POST /api/chat POST /api/pull POST /v1/chat/completions POST /v1/embeddings GET /v1/models
Why EWACS

One app. Two APIs. Your machine.

Point any Ollama or OpenAI client at EWACS and it just works — then add the things Ollama doesn't give you.

Ollama-compatible

The full /api/* surface — generate, chat, embeddings, tags, show, pull, create — with byte-for-byte streaming.

v1

OpenAI surface

/v1/chat/completions with SSE, plus completions, embeddings, and models. Your OpenAI SDK doesn't know the difference.

©

Licensed to ship

Every model in the marketplace is curated for a license that permits commercial use — gated, so you never run what you can't ship.

Runs on your machine

Your weights, your hardware, your network. The app does the inference locally on Windows and Linux; macOS best-effort.

One-click models

Sign in, pick a model from the catalog, and the app pulls and loads it for you — no manifests, no registries to wrangle.

Pluggable inference

Engines sit behind a trait, so the app can grow new backends and hardware support without changing the APIs your clients use.

Get started

Installed and chatting in a minute.

No build, no toolchain. Download the app, sign in, and pull a model.

1

Download & install

Grab the app for Windows or Linux and install it like any other desktop app.

2

Sign in for a token

Sign in with Google to get your marketplace token, then paste it into the app.

3

Pick a model & chat

Choose a model from the catalog; the app pulls, loads, and runs it locally. Start chatting.

Already running? Point your existing Ollama or OpenAI client at localhost:11434 — the /api/* and /v1/* endpoints work unchanged.

The marketplace

Pull models you're allowed to use.

We run a curated marketplace of open-source models whose licenses permit commercial use. We do the license homework so you don't have to: the license travels with every model, and a non-commercial license is a hard stop, not a footnote. You just sign in and pull — the app handles the rest.

1. sign in → token
2. pick "model:tag" in the app
3. license gate → commercial use?
4. verify digests → download & run locally
✗ non-commercial → 451, never served