Skip to content

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.

17.1 KB
Raw Blame
<!doctype html>
<html lang="en">
	<head>
		<script>/* three.ws theme boot: no-flash */(function(){try{var m=localStorage.getItem('twx_theme');var l=m==='auto'?(window.matchMedia&&window.matchMedia('(prefers-color-scheme: light)').matches):m==='light';document.documentElement.setAttribute('data-theme',l?'light':'dark');}catch(e){document.documentElement.setAttribute('data-theme','dark');}})();</script>
		<meta charset="utf-8" />
		<meta http-equiv="X-UA-Compatible" content="IE=edge" />
		<title data-i18n="gestures.page_title">Agent Gestures · three.ws</title>
		<meta name="viewport" content="width=device-width, initial-scale=1.0" />
		<meta name="description" content="Every gesture a three.ws agent can play: the slot vocabulary, the clip each slot resolves to, the skills that trigger it, and a builder for overriding any slot on your own agent." data-i18n-attr="content:gestures.meta_desc" />
		<meta name="theme-color" content="#000000" />
		<link rel="canonical" href="https://three.ws/gestures" />

		<meta property="og:type" content="website" />
		<meta property="og:site_name" content="three.ws" />
		<meta property="og:title" content="Agent Gestures · three.ws" data-i18n-attr="content:gestures.share_title" />
		<meta property="og:description" content="The agent gesture vocabulary: 19 slots, the clip each plays, the skills that trigger them, and a live override builder." data-i18n-attr="content:gestures.og_desc" />
		<meta property="og:url" content="https://three.ws/gestures" />
		<meta property="og:image" content="https://three.ws/api/page-og?s=main&amp;t=Agent+Gestures&amp;d=The+three.ws+agent+gesture+vocabulary%3A+every+animation+slot%2C+the+clip+it+plays%2C+the+skills+that+trigger+it%2C+and+a+builder+for+overriding+any+slot+on+your+own&amp;p=%2Fgestures" />
		<meta name="twitter:card" content="summary_large_image" />
		<meta name="twitter:site" content="@trythreews" />
		<meta name="twitter:title" content="Agent Gestures · three.ws" data-i18n-attr="content:gestures.share_title" />
		<meta name="twitter:description" content="19 gesture slots, previewed live, with the skills that trigger each one." data-i18n-attr="content:gestures.tw_desc" />
		<meta name="twitter:image" content="https://three.ws/api/page-og?s=main&amp;t=Agent+Gestures&amp;d=The+three.ws+agent+gesture+vocabulary%3A+every+animation+slot%2C+the+clip+it+plays%2C+the+skills+that+trigger+it%2C+and+a+builder+for+overriding+any+slot+on+your+own&amp;p=%2Fgestures" />

		<link rel="icon" href="/favicon.ico" sizes="any" />
		<link rel="icon" type="image/svg+xml" href="/favicon.svg" />
		<link rel="icon" type="image/png" sizes="32x32" href="/favicon-32x32.png" />
		<link rel="icon" type="image/png" sizes="16x16" href="/favicon-16x16.png" />
		<link rel="apple-touch-icon" href="/apple-touch-icon.png" />
		<link rel="manifest" href="/manifest.webmanifest" />
		<link rel="stylesheet" href="/style.css" />
		<link rel="stylesheet" href="/nav.css" />
		<link rel="stylesheet" href="/footer.css" />
		<link rel="stylesheet" href="/fonts/fonts.css" />
		<link rel="stylesheet" href="/gestures.css" />

		<script type="application/ld+json">{"@context":"https://schema.org","@type":"WebPage","name":"Agent Gestures","description":"The three.ws agent gesture vocabulary: every animation slot, the clip it resolves to, the skills that trigger it, and how to override one.","url":"https://three.ws/gestures","isPartOf":{"@type":"WebSite","name":"three.ws","url":"https://three.ws"}}</script>

		<meta name="robots" content="index, follow, max-image-preview:large, max-snippet:-1">
		<meta property="og:locale" content="en_US">
		<meta property="og:image:alt" content="Agent Gestures on three.ws">
	</head>
	<body class="gv-page">
		<header>
			<div id="nav-container"></div>
		</header>
		<script src="/nav.js"></script>

		<main>
			<section class="gv-hero">
				<div class="gv-hero-chip">
					<span class="gv-hero-dot" aria-hidden="true"></span>
					<span data-i18n="gestures.hero_chip">Agent body language</span>
				</div>
				<h1 data-i18n="gestures.hero_title">Agent gestures</h1>
				<p class="gv-hero-lede" data-i18n-html="gestures.hero_lede">
					An agent does not pick animation files. It names a <strong>slot</strong>, and the
					platform resolves that slot to a real clip. Skills name a slot indirectly, through a
					hint. This page is the whole vocabulary: what each slot plays, which skills trigger
					it, and how to point a slot at any other clip in the library.
				</p>
				<div class="gv-hero-actions">
					<a class="gv-btn gv-btn--primary" href="/choreograph" data-i18n="gestures.cta_choreograph">Sequence gestures into a routine</a>
					<a class="gv-btn gv-btn--secondary" href="/animations" data-i18n="gestures.cta_animations">Browse every clip</a>
					<a class="gv-btn gv-btn--ghost" href="/pose" data-i18n="gestures.cta_pose">Animation Studio</a>
					<a class="gv-btn gv-btn--ghost" href="/sonar" data-i18n="gestures.cta_sonar">Play these by hand, no camera</a>
					<a class="gv-btn gv-btn--ghost" href="/docs/animations" data-i18n="gestures.cta_docs">Developer reference</a>
				</div>
				<dl class="gv-stats" data-role="stats" aria-live="polite">
					<div class="gv-stat"><dt data-i18n="gestures.stat_slots">Slots</dt><dd data-role="stat-slots">&mdash;</dd></div>
					<div class="gv-stat"><dt data-i18n="gestures.stat_clips">Clips in library</dt><dd data-role="stat-clips">&mdash;</dd></div>
					<div class="gv-stat"><dt data-i18n="gestures.stat_hints">Skill hints</dt><dd data-role="stat-hints">&mdash;</dd></div>
					<div class="gv-stat"><dt data-i18n="gestures.stat_skills">Skills wired</dt><dd data-role="stat-skills">&mdash;</dd></div>
				</dl>
			</section>

			<section class="gv-body">
				<!-- Left rail: the live stage. Sticky on desktop so a card click is
				     always visible without scrolling back up. -->
				<aside class="gv-stage-col">
					<div class="gv-stage-card">
						<div class="gv-stage" data-role="stage">
							<div class="gv-stage-empty" data-role="stage-empty">
								<div class="gv-stage-empty-icon" aria-hidden="true">🎬</div>
								<p class="gv-stage-empty-title" data-i18n="gestures.stage_empty_title">Pick a gesture</p>
								<p class="gv-stage-empty-body" data-i18n="gestures.stage_empty_body">
									Choose any slot to play it on a live avatar. Nothing 3D loads until you do.
								</p>
							</div>
							<div class="gv-stage-loading" data-role="stage-loading" hidden>
								<div class="gv-spinner" aria-hidden="true"></div>
								<p data-i18n="gestures.stage_loading">Retargeting the clip&hellip;</p>
							</div>
							<div class="gv-stage-error" data-role="stage-error" hidden>
								<p class="gv-stage-error-title" data-role="stage-error-title" data-i18n="gestures.stage_error_title">That clip did not fit the preview avatar.</p>
								<p class="gv-stage-error-body" data-i18n-html="gestures.stage_error_body">
									The clip is fine: the shared preview rig could not bind it. Every clip is
									still previewable in the <a href="/animations">gallery</a>.
								</p>
								<button type="button" class="gv-btn gv-btn--secondary" data-role="stage-retry" data-i18n="gestures.stage_retry">Try again</button>
							</div>
						</div>
						<div class="gv-stage-meta">
							<div class="gv-stage-titles">
								<p class="gv-stage-slot" data-role="stage-slot" data-i18n="gestures.stage_slot_default">No gesture selected</p>
								<p class="gv-stage-clip" data-role="stage-clip">&nbsp;</p>
							</div>
							<div class="gv-stage-controls">
								<button type="button" class="gv-icon-btn" data-role="stage-replay" title="Replay" aria-label="Replay" disabled data-i18n-attr="title:gestures.replay_aria;aria-label:gestures.replay_aria">↻</button>
								<button type="button" class="gv-icon-btn" data-role="stage-pause" title="Pause" aria-label="Pause" disabled data-i18n-attr="title:gestures.pause_aria;aria-label:gestures.pause_aria">❙❙</button>
							</div>
						</div>
						<p class="gv-stage-note" data-role="stage-note" hidden></p>
						<p class="gv-sr-status" data-role="stage-status" role="status" aria-live="polite"></p>
					</div>

					<!-- Motion signature: measured from the clip's own keyframes by
					     scripts/build-motion-signatures.mjs, not authored by hand. -->
					<div class="gv-sig" data-role="signature" hidden>
						<h2 class="gv-sig-title" data-i18n="gestures.sig_title">What this clip actually does</h2>
						<p class="gv-sig-sentence" data-role="signature-sentence"></p>
						<div class="gv-fp gv-fp--full" data-role="signature-bar"></div>
						<dl class="gv-sig-stats" data-role="signature-stats"></dl>
						<div class="gv-sig-badges" data-role="signature-badges"></div>
					</div>

					<!-- The verdict: can this clip do the job this slot is for. -->
					<div class="gv-fit" data-role="fit" hidden>
						<span class="gv-fit-icon" aria-hidden="true"></span>
						<p data-role="fit-text"></p>
					</div>

					<div class="gv-similar" data-role="similar" hidden>
						<h2 class="gv-similar-title" data-i18n="gestures.similar_title">Moves like this one</h2>
						<p class="gv-similar-lede" data-i18n="gestures.similar_lede">
							Nearest clips in the measured motion space, not by name.
						</p>
						<div class="gv-chips" data-role="similar-list"></div>
					</div>

					<!-- Override builder. The output is exactly the shape an agent's
					     meta.edits.animations takes, so it is copy-paste usable. -->
					<div class="gv-override" data-role="override" hidden>
						<h2 class="gv-override-title" data-i18n="gestures.override_title">Override this slot</h2>
						<p class="gv-override-lede" data-i18n-html="gestures.override_lede">
							Point <code data-role="override-slot">slot</code> at any clip in the library, then
							save it straight to an agent you own.
						</p>
						<label class="gv-field">
							<span class="gv-field-label" data-i18n="gestures.field_clip_label">Clip</span>
							<select class="gv-select" data-role="override-select"></select>
						</label>
						<!-- Apply straight to an agent you own. The alternative is copying
						     the JSON into the agent editor by hand, which is four more
						     steps for the same result. -->
						<div class="gv-apply" data-role="apply" hidden>
							<label class="gv-field">
								<span class="gv-field-label" data-i18n="gestures.field_apply_label">Apply to</span>
								<select class="gv-select" data-role="apply-select"></select>
							</label>
							<div class="gv-override-actions">
								<button type="button" class="gv-btn gv-btn--primary" data-role="apply-save" data-i18n="gestures.apply_save">Save to agent</button>
								<button type="button" class="gv-btn gv-btn--ghost" data-role="override-copy" data-i18n="gestures.copy_json">Copy JSON</button>
							</div>
							<p class="gv-apply-status" data-role="apply-status" role="status" aria-live="polite"></p>
						</div>

						<div class="gv-apply" data-role="apply-signedout" hidden>
							<div class="gv-override-actions">
								<a class="gv-btn gv-btn--primary" href="/login?next=/gestures" data-role="apply-signin" data-i18n="gestures.apply_signin">Sign in to save this to an agent</a>
								<button type="button" class="gv-btn gv-btn--ghost" data-role="override-copy-2" data-i18n="gestures.copy_json">Copy JSON</button>
							</div>
						</div>

						<details class="gv-json">
							<summary data-i18n="gestures.show_json">Show the JSON</summary>
							<pre class="gv-code" data-role="override-code"><code></code></pre>
						</details>

						<div class="gv-override-actions">
							<button type="button" class="gv-btn gv-btn--ghost" data-role="override-reset" data-i18n="gestures.override_reset">Reset to default</button>
							<!-- Carries the staged slot and clip straight into the sequencer,
							     so "this gesture is good" turns into a routine without
							     re-picking it there. Href is rewritten on every selection. -->
							<a class="gv-btn gv-btn--ghost" href="/choreograph" data-role="override-sequence" data-i18n="gestures.override_sequence">Use in a routine</a>
						</div>
					</div>
				</aside>

				<!-- Right rail: the vocabulary. -->
				<div class="gv-main-col">
					<div class="gv-section-head">
						<h2 id="slots" data-i18n="gestures.slots_h">The slot vocabulary</h2>
						<p data-i18n="gestures.slots_desc">
							Fixed names an agent can override. A slot with no override resolves to the clip
							shown on its card.
						</p>
					</div>

					<div class="gv-grid" data-role="grid" role="list" aria-busy="true">
						<div class="gv-card gv-card--skeleton" aria-hidden="true"></div>
						<div class="gv-card gv-card--skeleton" aria-hidden="true"></div>
						<div class="gv-card gv-card--skeleton" aria-hidden="true"></div>
						<div class="gv-card gv-card--skeleton" aria-hidden="true"></div>
						<div class="gv-card gv-card--skeleton" aria-hidden="true"></div>
						<div class="gv-card gv-card--skeleton" aria-hidden="true"></div>
					</div>

					<div class="gv-error" data-role="grid-error" hidden>
						<p class="gv-error-title" data-i18n="gestures.grid_error_title">The clip library did not load.</p>
						<p class="gv-error-body" data-role="grid-error-detail"></p>
						<button type="button" class="gv-btn gv-btn--secondary" data-role="grid-retry" data-i18n="gestures.retry_btn">Retry</button>
					</div>

					<p class="gv-grid-hint" data-i18n-html="gestures.grid_hint">
						<kbd>↑</kbd> <kbd>↓</kbd> <kbd>←</kbd> <kbd>→</kbd> move · <kbd>Enter</kbd> play ·
						<kbd>Esc</kbd> stop
					</p>

					<div class="gv-section-head">
						<h2 id="hints" data-i18n="gestures.hints_h">What triggers a gesture</h2>
						<p data-i18n="gestures.hints_desc">
							A skill declares a hint when it starts. The hint resolves to a slot, the slot
							resolves to a clip, and the avatar plays it. Every row below is generated from
							the shipped skill definitions.
						</p>
					</div>

					<div class="gv-hints" data-role="hints">
						<div class="gv-hints-loading" data-i18n="gestures.hints_loading">Loading the skill map&hellip;</div>
					</div>

					<div class="gv-section-head">
						<h2 id="atlas" data-i18n="gestures.atlas_h">The motion atlas</h2>
						<p data-i18n="gestures.atlas_desc">
							Every clip in the library, measured from its own keyframes and plotted
							by how energetic it is and how much of that energy is above the waist.
							Calm upper-body gestures gather top-left; explosive full-body dances and
							runs sit bottom-right. Colour is the region leading the movement. Click
							any point to play it, or to try it in the slot you have selected.
						</p>
					</div>

					<div class="gv-atlas-tools">
						<div class="gv-pills" data-role="atlas-filters" role="group" aria-label="Filter the atlas" data-i18n-attr="aria-label:gestures.atlas_filters_aria"></div>
						<span class="gv-atlas-count" data-role="atlas-count" aria-live="polite"></span>
					</div>
					<div class="gv-atlas" data-role="atlas"></div>
					<div class="gv-atlas-legend">
						<span class="gv-fp-key" data-i18n-html="gestures.legend_head"><i data-region="head"></i>Head</span>
						<span class="gv-fp-key" data-i18n-html="gestures.legend_arms"><i data-region="arms"></i>Arms</span>
						<span class="gv-fp-key" data-i18n-html="gestures.legend_torso"><i data-region="torso"></i>Torso</span>
						<span class="gv-fp-key" data-i18n-html="gestures.legend_hips"><i data-region="root"></i>Hips</span>
						<span class="gv-fp-key" data-i18n-html="gestures.legend_legs"><i data-region="legs"></i>Legs</span>
					</div>
					<div class="gv-error" data-role="atlas-error" hidden>
						<p class="gv-error-title" data-i18n="gestures.atlas_error_title">The motion index did not load.</p>
						<p class="gv-error-body" data-role="atlas-error-detail"></p>
						<button type="button" class="gv-btn gv-btn--secondary" data-role="atlas-retry" data-i18n="gestures.retry_btn">Retry</button>
					</div>

					<div class="gv-section-head">
						<h2 id="layers" data-i18n="gestures.layers_h">Player gestures on /walk</h2>
						<p data-i18n-html="gestures.layers_desc">
							The walk state machine keeps a second, player-driven gesture library with a layer
							per gesture: <code>upper</code> gestures overlay locomotion, so an avatar waves or
							shrugs while it keeps walking; <code>full</code> gestures take the whole body and
							suppress movement until they end.
						</p>
					</div>
					<div class="gv-layers" data-role="layers"></div>

					<div class="gv-footnote">
						<p data-i18n-html="gestures.footnote">
							Slots, hints and clips are held together mechanically:
							<code>tests/animation-slots.test.js</code> scans every hint declared under
							<code>src/</code> and fails the build if one resolves to no slot or to a clip that
							is not baked in <code>public/animations/manifest.json</code>. The full developer
							reference is <a href="/docs/animations">docs/animations</a>, and the walkthrough is
							<a href="/tutorials/animate-your-agent">Animate your agent</a>.
						</p>
					</div>
				</div>
			</section>
		</main>

		<div id="footer-container"></div>
		<script src="/footer.js" defer></script>

		<script type="module" src="/src/gestures-page.js"></script>
		<!-- Runtime i18n: detects locale, swaps annotated copy, mounts <lang-switcher>. -->
		<script type="module" src="/i18n.js"></script>
	</body>
</html>