Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
Indie Creator — "Give me a 3D version of myself I can put anywhere"
Every scenario below is an example workflow, not a real customer. Features and routes are re-confirmed against
README.md.
Who this is for
You're a solo creator — a streamer, an illustrator, a small-game maker, a personal-brand builder. You want a 3D avatar that is you (or an original character), that can talk, and that you can drop onto your site, your link-in-bio, or your store page. You don't run a render farm, you don't write WebGL, and you don't want to hand your likeness to a tool that uploads it to a server you can't see.
The problem, concretely
Getting a usable, riggable 3D avatar today means one of: hiring a 3D artist (slow, expensive), learning Blender (weeks), or wrestling a generic avatar maker that produces something that looks nothing like you. Then, even if you get a GLB, embedding it is a second project — most viewers are heavyweight, framework-locked, or can't talk. The result is that most creators never ship a 3D presence at all.
How three.ws solves it
Three real, shipped features collapse that whole chain into minutes:
- Selfie → 3D —
/scanand/create/selfieturn a photo into a rigged 3D avatar. Rendering and capture run in the browser — drag-and-drop GLB handling is fully client-side (no server upload of the model itself). - Text → 3D (Forge) —
/forgeis the in-house text-to-3D pipeline: describe a character in words and get a 3D model back, when you'd rather invent an original than scan your face. - Embed anywhere — the
<agent-3d>web component drops onto any HTML page with one script tag, no framework. Add abrain=attribute and the avatar talks. Widget Studio and the Embed Editor generate the snippet for you.
Together: your face (or your prompt) becomes a model, the model becomes a talking agent, and the agent becomes a copy-paste embed. Every step is a real route.
Example workflow (hypothetical)
Imagine an indie illustrator, "Procyon," who sells art prints from a one-page site and wants a 3D version of themselves greeting visitors. Here's the path they'd take.
- Procyon opens
/scanon their phone, follows the capture UX, and gets a rigged avatar — or, wanting a stylized fox mascot instead of their own face, opens/forgeand types "a friendly anthropomorphic fox in a denim jacket, soft studio lighting." Forge returns a GLB. - They open the Embed Editor, pick the avatar, choose an idle animation and a transparent background, and frame it.
- They set a brain and personality inline, copying the snippet the editor produces:
<script type="module" src="https://three.ws/agent-3d/1.5.2/agent-3d.js"></script> <agent-3d body="https://three.ws/avatars/fox.glb" brain="claude-sonnet-4-6" name="Procyon" instructions="You are Procyon's studio fox. Greet visitors warmly, talk about the art prints, and wave when someone says hi." mode="floating" position="bottom-right" width="320px" height="420px" ></agent-3d> - They paste that into their site's HTML (works on a static site, Ghost, WordPress, Webflow). The floating fox now talks to every visitor.
- Deliverable: a talking 3D mascot pinned to the corner of their store, reachable by every visitor — built from a prompt and a copy-paste, with no 3D software and no build step.
What you get
A real, embedded, conversational 3D avatar that loads in any modern browser. The likeness comes from your actual selfie (or your actual prompt), not a fabricated demo. It speaks via Claude when you set a brain, falls back to text where speech isn't available, and renders with full PBR lighting. Honest scope note: avatar quality depends on your input photo/prompt, and the selfie reconstruction backend is a Phase-1 capability still hardening — the capture UX and embedding are live today.
Next step / CTA
- Start:
/scan(selfie) or/forge(text) → Embed Editor. - Learn: Build Your First Agent · Embed on Your Website.
- Social spotlight angle (G03): "From one selfie to a talking 3D you, embedded in your bio — no Blender."
[REAL CASE STUDY — fill on consent: an indie creator who shipped an embedded agent and the response it got.]