← Blog

Showcasing the three.ws × IBM Partnership

three.ws and IBM are partners, and we shared a live demo publicly. We showed a persistent, on-chain 3D agent embedded into a page as easily as a YouTube video, thinking on IBM watsonx.ai — and IBM's official account replied in the open:

"We're super excited about this."

This post is the longer version of that demo: what we showed and how an agent thinks on IBM watsonx.ai. A note before we start: the live demos linked here are independent tools we built for developers to explore IBM Granite and build their own — not official partnership deliverables. Our formal partnership work with IBM is being built on the IBM platform and is not yet public.

What we showed

The demo made one thing concrete: embedding a persistent, on-chain 3D agent into any platform should be as easy as embedding a YouTube video. One <agent-3d> tag, dropped into a page, deploys a full agent — no install, no plugin, no native app. It renders in any modern browser over WebGL.

What lands on the page is not a chat box. It's a character with a face, a voice, memory, emotion, and an on-chain identity that persists across platforms. The goal of that frictionless deployment is simple: onboard the next millions of people to embodied AI without asking any of them to download anything.

Powered by IBM watsonx.ai

Behind the 3D body is a brain — and that brain can run on IBM watsonx.ai. The integration is real, in production, and open source. There is no mock path.

Two layers connect three.ws to watsonx:

Both layers run on your own IBM Cloud credentials. An API key is exchanged for a short-lived IAM bearer token, cached until just before expiry, and every inference call is scoped to your watsonx.ai project — no three.ws intermediary sits between your agent and IBM. The result: a 3D agent embedded on any page can reason on IBM Granite, running inside your own watsonx.ai project, governed by your IBM Cloud account.

A formal IBM partnership

This is a formal partnership: three.ws is a partner in IBM's Business Partner program. As we shared earlier, that partnership brings together the open-source, browser-native, decentralized on-chain identity of three.ws with IBM's leadership in enterprise AI — and gives three.ws an integration path into IBM's AI and cloud portfolio, plus a route toward enterprise deployments in regulated industries.

Why enterprise AI needs a body

Enterprise AI deployments today are mostly headless — text in, text out, embedded in dashboards and chat windows that feel like forms. three.ws is the layer that gives those agents a persistent presence: a face that reacts to data, a voice, a body, and an identity that survives across surfaces. Concretely, it brings enterprise AI:

See it live

These demos are tools we built for developers, not partnership deliverables. Separate from the partnership itself, we built a suite of live demos that put IBM's publicly available Granite models inside an embodied, on-chain agent — independent showcases for developers to explore and build their own tools on. They are not official IBM partnership deliverables, not IBM products, and not endorsed by IBM. Each one runs on real watsonx.ai inference:

Start at the three.ws × IBM hub.

Try it

The watsonx integration is open source and runnable today. Point it at your own IBM Cloud account:

WATSONX_API_KEY=... WATSONX_PROJECT_ID=... npx @three-ws/ibm-watsonx-mcp

You'll need an IBM Cloud account with watsonx.ai provisioned (free tier available), an IBM Cloud API key, and your watsonx.ai project id.

Links

Related


three.ws is an independent IBM Business Partner. The demos at three.ws/ibm and the @three-ws/ibm-watsonx-mcp package are community-built tools for developers and are not IBM products, official IBM partnership deliverables, or endorsed by IBM. Our formal partnership work with IBM is being built on the IBM platform and is not yet public. IBM, watsonx, watsonx.ai, IBM Cloud, and Granite are trademarks of International Business Machines Corporation.

← All posts