Skip to content

The ultimate harness for coding agents and beyond. All your agents. All your machines. One command center. Start with code, then follow your curiosity and build across disciplines: CAD, circuits, robots, games and music.

Rebuild the README for developers (#237)

Dee3d8011ac5348
diff --git a/.github/assets/coding/README.md b/.github/assets/coding/README.md
new file mode 100644
index 000000000..e0c0f0b89
--- /dev/null
+++ b/.github/assets/coding/README.md
@@ -0,0 +1,13 @@
+# Coding workspace preview
+
+`workspace.gif` is a looping preview of the [original Harness coding video](https://cdn.autonomous.ai/development/ecm/260910/Thumb-harness-app.mp4),
+provided for the README. It keeps the complete clip at its original speed and aspect ratio,
+at 1280 × 720 and 12 frames per second. The README links the animation to the original MP4.
+
+Regenerate after downloading the source as `harness-coding-demo.mp4`:
+
+```sh
+ffmpeg -i harness-coding-demo.mp4 \
+ -filter_complex '[0:v]fps=12,scale=1280:-1:flags=lanczos,split[a][b];[a]palettegen=stats_mode=diff[p];[b][p]paletteuse=dither=bayer:bayer_scale=4:diff_mode=rectangle' \
+ -loop 0 .github/assets/coding/workspace.gif
+```
diff --git a/.github/assets/coding/workspace.gif b/.github/assets/coding/workspace.gif
new file mode 100644
index 000000000..e86855818
Binary files /dev/null and b/.github/assets/coding/workspace.gif differ
diff --git a/.github/assets/readme/.gitignore b/.github/assets/readme/.gitignore
new file mode 100644
index 000000000..ceddaa37f
--- /dev/null
+++ b/.github/assets/readme/.gitignore
@@ -0,0 +1 @@
+.cache/
diff --git a/.github/assets/readme/README.md b/.github/assets/readme/README.md
new file mode 100644
index 000000000..68c485b35
--- /dev/null
+++ b/.github/assets/readme/README.md
@@ -0,0 +1,33 @@
+# README animations
+
+| File | What it shows | Made by |
+|---|---|---|
+| `agents.gif` | Four agents on three machines, working at once | `render.mjs` |
+| `machines.gif` | A new box set up in four commands, linked, and running an agent | `render.mjs` |
+| `keyboard.gif` | Open, zoom, answer a waiting agent and split, keyboard only | `render.mjs` |
+| `e2ee.gif` | A session beside the ciphertext the relay carries | `render.mjs` |
+| `beyond/<harness>.gif` | One recorded hands-on session per harness, full length at real speed | `build.py` |
+| `connect/*.gif` | Direct, Cloudflare and relay paths, from autonomous.ai/harness-app | `site.mjs` |
+| `device.gif` | The Harness device video from autonomous.ai/harness-device | `site.mjs` |
+
+The first four are scripted scenes in `scenes/index.html`, drawn in the desktop app's design:
+the tab bar, pane headers with machine, project and branch, and the fzf-style command box.
+Open the file with `#agents`, `#machines`, `#keyboard` or `#e2ee` to watch a scene live.
+Shortcuts, commands and dialog text follow `docs/keyboard.md`, `docs/cli.md` and the app's strings;
+encryption facts follow `docs/architecture.md`. Keep them that way when editing a scene.
+
+The eight `beyond/` GIFs convert the unedited recordings in `docs/images/*-demo.mp4`.
+
+Regenerate from the repository root:
+
+```sh
+(cd store/tools/experience-tests && npm ci) # once, for playwright-core
+node .github/assets/readme/render.mjs # all four scenes, or name one
+node .github/assets/readme/render.mjs --stills 2,6 keyboard # review stills in .cache/
+python3 .github/assets/readme/build.py # beyond/*.gif
+node .github/assets/readme/site.mjs # connect/*.gif and device.gif
+```
+
+`render.mjs` needs Google Chrome and FFmpeg. It renders at twice the scene size and scales down,
+so text stays sharp at README width. `build.py` needs FFmpeg. It writes 800 px GIFs at 8 fps, 6 fps for the busier
+Strudel recording, so each stays under 7 MiB.
diff --git a/.github/assets/readme/agents.gif b/.github/assets/readme/agents.gif
new file mode 100644
index 000000000..5f790b32d
Binary files /dev/null and b/.github/assets/readme/agents.gif differ
diff --git a/.github/assets/readme/beyond/blender.gif b/.github/assets/readme/beyond/blender.gif
new file mode 100644
index 000000000..5fec5e7e2
Binary files /dev/null and b/.github/assets/readme/beyond/blender.gif differ
diff --git a/.github/assets/readme/beyond/circuitjs.gif b/.github/assets/readme/beyond/circuitjs.gif
new file mode 100644
index 000000000..384452f8c
Binary files /dev/null and b/.github/assets/readme/beyond/circuitjs.gif differ
diff --git a/.github/assets/readme/beyond/godogen.gif b/.github/assets/readme/beyond/godogen.gif
new file mode 100644
index 000000000..33a77d802
Binary files /dev/null and b/.github/assets/readme/beyond/godogen.gif differ
diff --git a/.github/assets/readme/beyond/jev-sheets.gif b/.github/assets/readme/beyond/jev-sheets.gif
new file mode 100644
index 000000000..0dd4dc865
Binary files /dev/null and b/.github/assets/readme/beyond/jev-sheets.gif differ
diff --git a/.github/assets/readme/beyond/mujoco.gif b/.github/assets/readme/beyond/mujoco.gif
new file mode 100644
index 000000000..5c3bde733
Binary files /dev/null and b/.github/assets/readme/beyond/mujoco.gif differ
diff --git a/.github/assets/readme/beyond/rdkit.gif b/.github/assets/readme/beyond/rdkit.gif
new file mode 100644
index 000000000..bde212d9b
Binary files /dev/null and b/.github/assets/readme/beyond/rdkit.gif differ
diff --git a/.github/assets/readme/beyond/strudel.gif b/.github/assets/readme/beyond/strudel.gif
new file mode 100644
index 000000000..bbeaf6cf0
Binary files /dev/null and b/.github/assets/readme/beyond/strudel.gif differ
diff --git a/.github/assets/readme/beyond/typst.gif b/.github/assets/readme/beyond/typst.gif
new file mode 100644
index 000000000..8b0f209e8
Binary files /dev/null and b/.github/assets/readme/beyond/typst.gif differ
diff --git a/.github/assets/readme/build.py b/.github/assets/readme/build.py
new file mode 100644
index 000000000..14844173c
--- /dev/null
+++ b/.github/assets/readme/build.py
@@ -0,0 +1,33 @@
+#!/usr/bin/env python3
+"""Convert the eight hands-on recordings into README GIFs, one per harness.
+
+Run from the repository root: python3 .github/assets/readme/build.py
+Needs FFmpeg. Sources are the unedited recordings in docs/images/*-demo.mp4;
+output goes to .github/assets/readme/beyond/<harness>.gif at full length and real speed.
+"""
+import os, subprocess
+
+ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), '..', '..', '..'))
+OUT = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'beyond')
+WIDTH = 800
+
+GIFS = [ # harness, recording, frames per second
+ ('blender', 'blender-shape-lab-demo.mp4', 8),
+ ('circuitjs', 'scope-lab-demo.mp4', 8),
+ ('mujoco', 'mujoco-what-if-demo.mp4', 8),
+ ('godogen', 'godogen-rewind-demo.mp4', 8),
+ ('strudel', 'strudel-live-take-demo.mp4', 6), # busiest screen; fewer frames keep it light
+ ('rdkit', 'rdkit-bond-scan-demo.mp4', 8),
+ ('typst', 'doc-review-demo.mp4', 8),
+ ('jev-sheets', 'question-lab-demo.mp4', 8),
+]
+
+os.makedirs(OUT, exist_ok=True)
+for name, src, fps in GIFS:
+ out = os.path.join(OUT, f'{name}.gif')
+ vf = (f'fps={fps},scale={WIDTH}:-2:flags=lanczos,split[a][b];'
+ f'[a]palettegen=max_colors=192:stats_mode=diff[p];[b][p]paletteuse=dither=none:diff_mode=rectangle')
+ # Skip the recording's first moments, where the page is still loading.
+ subprocess.run(['ffmpeg', '-v', 'error', '-y', '-ss', '0.4', '-i', os.path.join(ROOT, 'docs/images', src),
+ '-filter_complex', vf, '-loop', '0', out], check=True)
+ print(f'{name}.gif {os.path.getsize(out) / 1048576:.1f} MiB')
diff --git a/.github/assets/readme/connect/cloudflare.gif b/.github/assets/readme/connect/cloudflare.gif
new file mode 100644
index 000000000..199fdc05b
Binary files /dev/null and b/.github/assets/readme/connect/cloudflare.gif differ
diff --git a/.github/assets/readme/connect/direct.gif b/.github/assets/readme/connect/direct.gif
new file mode 100644
index 000000000..346d22906
Binary files /dev/null and b/.github/assets/readme/connect/direct.gif differ
diff --git a/.github/assets/readme/connect/relay.gif b/.github/assets/readme/connect/relay.gif
new file mode 100644
index 000000000..59cfc8c44
Binary files /dev/null and b/.github/assets/readme/connect/relay.gif differ
diff --git a/.github/assets/readme/device.gif b/.github/assets/readme/device.gif
new file mode 100644
index 000000000..880708e5c
Binary files /dev/null and b/.github/assets/readme/device.gif differ
diff --git a/.github/assets/readme/e2ee.gif b/.github/assets/readme/e2ee.gif
new file mode 100644
index 000000000..928862ce6
Binary files /dev/null and b/.github/assets/readme/e2ee.gif differ
diff --git a/.github/assets/readme/keyboard.gif b/.github/assets/readme/keyboard.gif
new file mode 100644
index 000000000..f593b9f46
Binary files /dev/null and b/.github/assets/readme/keyboard.gif differ
diff --git a/.github/assets/readme/machines.gif b/.github/assets/readme/machines.gif
new file mode 100644
index 000000000..35ccfb7a0
Binary files /dev/null and b/.github/assets/readme/machines.gif differ
diff --git a/.github/assets/readme/render.mjs b/.github/assets/readme/render.mjs
new file mode 100644
index 000000000..503e02ba9
--- /dev/null
+++ b/.github/assets/readme/render.mjs
@@ -0,0 +1,60 @@
+#!/usr/bin/env node
+// Render the README feature GIFs from scenes/index.html.
+//
+// node .github/assets/readme/render.mjs all scenes
+// node .github/assets/readme/render.mjs keyboard one scene
+// node .github/assets/readme/render.mjs --stills 2,6 PNG stills at those seconds, for review
+//
+// Needs Google Chrome, FFmpeg, and playwright-core from store/tools/experience-tests
+// (run `npm ci` there once).
+import { execFileSync } from 'node:child_process';
+import { mkdirSync, rmSync, statSync } from 'node:fs';
+import { dirname, join, resolve } from 'node:path';
+import { fileURLToPath, pathToFileURL } from 'node:url';
+
+const here = dirname(fileURLToPath(import.meta.url));
+const root = resolve(here, '../../..');
+const { chromium } = await import(pathToFileURL(join(root, 'store/tools/experience-tests/node_modules/playwright-core/index.mjs')).href);
+
+const args = process.argv.slice(2);
+const stillsAt = args.includes('--stills') ? args[args.indexOf('--stills') + 1].split(',').map(Number) : null;
+const names = args.filter((a, i) => !a.startsWith('--') && args[i - 1] !== '--stills');
+const FPS = 15, OUT_W = 1280, FADE = 0.4;
+const work = process.env.README_RENDER_TMP || join(here, '.cache', 'frames');
+
+const browser = await chromium.launch({ channel: 'chrome' });
+const page = await browser.newPage({ viewport: { width: 960, height: 600 }, deviceScaleFactor: 2 });
+await page.goto(pathToFileURL(join(here, 'scenes', 'index.html')).href);
+const scenes = names.length ? names : await page.evaluate(() => Object.keys(window.SCENES));
+
+async function frame(name, t, path) {
+ await page.evaluate(([n, s]) => window.draw(n, s), [name, t]);
+ await page.evaluate(() => Promise.all([...document.images].map(i => i.decode().catch(() => {}))));
+ await page.screenshot({ path });
+}
+
+for (const name of scenes) {
+ const dur = await page.evaluate(n => window.SCENES[n].dur, name);
+ if (stillsAt) {
+ mkdirSync(join(work, 'stills'), { recursive: true });
+ for (const t of stillsAt) await frame(name, t, join(work, 'stills', `${name}-${t}.png`));
+ continue;
+ }
+ const dir = join(work, name);
+ rmSync(dir, { recursive: true, force: true });
+ mkdirSync(dir, { recursive: true });
+ const n = Math.round(dur * FPS);
+ for (let i = 0; i < n; i++) await frame(name, i / FPS, join(dir, `${String(i).padStart(5, '0')}.png`));
+ // Loop cleanly: the last FADE seconds dissolve back into the first frame.
+ const gif = join(here, `${name}.gif`);
+ const off = (dur - FADE).toFixed(3);
+ execFileSync('ffmpeg', ['-v', 'error', '-y',
+ '-framerate', String(FPS), '-i', join(dir, '%05d.png'),
+ '-loop', '1', '-framerate', String(FPS), '-t', String(FADE), '-i', join(dir, '00000.png'),
+ '-filter_complex',
+ `[0][1]xfade=transition=fade:duration=${FADE}:offset=${off},scale=${OUT_W}:-2:flags=lanczos,split[a][b];` +
+ `[a]palettegen=max_colors=256:stats_mode=full[p];[b][p]paletteuse=dither=none`,
+ '-loop', '0', gif]);
+ console.log(`${name}.gif ${n} frames ${dur}s ${(statSync(gif).size / 1048576).toFixed(2)} MiB`);
+}
+await browser.close();
diff --git a/.github/assets/readme/scenes/index.html b/.github/assets/readme/scenes/index.html
new file mode 100644
index 000000000..b07c4daeb
--- /dev/null
+++ b/.github/assets/readme/scenes/index.html
@@ -0,0 +1,439 @@
+<!doctype html>
+<html>
+<head>
+<meta charset="utf-8">
+<title>README scenes</title>
+<!--
+ Deterministic scenes for the README GIFs, drawn in the Harness desktop design language:
+ the tab bar, pane headers with machine, project and branch, and the fzf-style command box.
+ render.mjs calls draw(scene, seconds) for every frame and screenshots the page.
+ Open this file in a browser and add #agents, #machines, #keyboard or #e2ee to preview a scene.
+-->
+<style>
+:root {
+ --bg: #181818; --panel: #141414; --line: #2b2b2b; --tab: #1c1c1c; --box: #101113; --row: #232428;
+ --fg: #e8e8e8; --dim: #8b8b8b; --dim2: #5d5d5d; --accent: #bdcbdc;
+ --green: #7ee787; --orange: #e8845c; --blue: #79c0ff; --yellow: #e3b341; --violet: #d2a8ff; --red: #ff7b72;
+}
+* { box-sizing: border-box; margin: 0; padding: 0; }
+html, body { width: 960px; height: 600px; overflow: hidden; background: var(--bg); color: var(--fg);
+ font: 12.5px/1.55 ui-monospace, "SF Mono", Menlo, monospace; -webkit-font-smoothing: antialiased; }
+#app { position: absolute; inset: 0; }
+.dim { color: var(--dim); } .dim2 { color: var(--dim2); } .g { color: var(--green); } .o { color: var(--orange); }
+.b { color: var(--blue); } .y { color: var(--yellow); } .v { color: var(--violet); } .r { color: var(--red); }
+.w { color: #fff; font-weight: 600; }
+
+/* tab bar */
+.tabbar { position: absolute; left: 0; top: 0; right: 0; height: 34px; background: var(--tab);
+ border-bottom: 1px solid #0e0e0e; display: flex; align-items: center; gap: 12px; padding: 0 12px; }
+.lights { display: flex; gap: 8px; margin-right: 4px; }
+.lights i { width: 12px; height: 12px; border-radius: 50%; display: block; }
+.lights i:nth-child(1) { background: #ff5f57; } .lights i:nth-child(2) { background: #febc2e; } .lights i:nth-child(3) { background: #28c840; }
+.bell { position: relative; color: var(--dim); display: flex; }
+.bell .badge { position: absolute; right: -7px; top: -5px; min-width: 14px; height: 14px; border-radius: 7px;
+ background: var(--orange); color: #111; font-size: 9.5px; font-weight: 700; line-height: 14px; text-align: center; }
+.tab { display: flex; align-items: center; gap: 8px; height: 24px; padding: 0 12px; border-radius: 6px;
+ background: #282828; color: var(--fg); }
+.plus { color: var(--dim); font-size: 15px; }
+.store { margin-left: auto; display: flex; align-items: center; gap: 6px; height: 22px; padding: 0 11px;
+ border-radius: 11px; background: #262626; color: var(--fg); font-size: 11.5px; }
+
+/* panes */
+.pane { position: absolute; background: var(--panel); border: 1px solid var(--line); border-radius: 6px; overflow: hidden; }
+.pane.focus { border-color: var(--accent); box-shadow: 0 0 0 1px rgba(189, 203, 220, .25); }
+.pane.new { animation: none; }
+.ph { height: 27px; display: flex; align-items: center; gap: 7px; padding: 0 10px; border-bottom: 1px solid #1e1e1e;
+ white-space: nowrap; overflow: hidden; }
+.ph img { width: 14px; height: 14px; border-radius: 3px; }
+.ph b { font-weight: 650; color: #fff; }
+.ph .meta { margin-left: auto; display: flex; gap: 11px; color: var(--dim); font-size: 11.5px; }
+.ph .meta span { display: flex; align-items: center; gap: 4px; }
+.ph .meta .pj, .ph .meta .br { color: var(--green); }
+.pt { padding: 8px 11px; white-space: pre; overflow: hidden; }
+.cur { display: inline-block; width: 7.5px; height: 15px; background: #e8e8e8; vertical-align: -3px; }
+.cur.off { opacity: 0; }
+.ask { margin-top: 6px; border: 1px solid #3a3a3a; border-radius: 5px; padding: 6px 9px; }
+
+/* command box */
+.box { position: absolute; left: 0; right: 0; bottom: 0; background: var(--box); border-top: 1px solid #2a2a2a; }
+.box .head { display: flex; justify-content: space-between; padding: 8px 14px 4px; color: var(--dim); font-size: 11.5px; }
+.box .field { display: flex; padding: 1px 14px; }
+.box .field .k { width: 78px; color: var(--dim); }
+.box .field.task { margin-top: 5px; padding-top: 5px; padding-bottom: 5px; }
+.box .field.task .v { color: #fff; }
+.box .foot { display: flex; gap: 16px; padding: 7px 14px 8px; margin-top: 6px; border-top: 1px solid #1d1e21;
+ color: var(--dim2); font-size: 11px; }
+.box .foot b { color: var(--dim); font-weight: 600; margin-right: 5px; }
+.box .list { display: flex; }
+.box .rows { flex: 1; padding: 0 6px; display: flex; flex-direction: column; justify-content: flex-end; }
+.box .row { display: flex; gap: 12px; align-items: center; height: 23px; padding: 0 8px; border-radius: 4px; white-space: nowrap; }
+.box .row.sel { background: var(--row); }
+.box .row .t { color: #fff; font-weight: 600; }
+.box .row .e { color: var(--dim); }
+.box .row .m { color: var(--dim); }
+.box .row .pj { color: var(--green); }
+.box .row .ret { margin-left: auto; color: var(--dim); }
+.box .preview { width: 300px; border-left: 1px solid #1d1e21; padding: 2px 12px; white-space: pre; color: var(--dim); font-size: 11.5px; }
+.box .preview .t { color: #fff; font-weight: 600; }
+.box .query { padding: 7px 14px 0; color: #fff; }
+
+/* keystroke overlay */
+.hud { position: absolute; left: 50%; transform: translateX(-50%); display: flex; gap: 6px; padding: 8px; border-radius: 12px;
+ background: rgba(30, 32, 36, .92); border: 1px solid #3a3d44; box-shadow: 0 8px 30px rgba(0, 0, 0, .5); }
+.hud kbd { min-width: 38px; height: 38px; padding: 0 10px; border-radius: 8px; background: #2d3036; border: 1px solid #4a4e57;
+ border-bottom-width: 3px; color: #fff; font: 600 19px/34px -apple-system, "SF Pro Text", system-ui; text-align: center; }
+.hud .label { font: 500 13px/38px -apple-system, "SF Pro Text", system-ui; color: #c9d1d9; padding: 0 6px 0 4px; }
+
+/* other machine's terminal */
+.dimmer { position: absolute; inset: 0; background: rgba(0, 0, 0, .55); }
+.term { position: absolute; left: 150px; top: 120px; width: 660px; height: 300px; border-radius: 10px; overflow: hidden;
+ background: #0b0c0e; border: 1px solid #3a3a3a; box-shadow: 0 30px 80px rgba(0, 0, 0, .7); }
+.term .bar { height: 30px; background: #1f2023; display: flex; align-items: center; padding: 0 12px; gap: 8px; color: var(--dim); font-size: 11.5px; }
+.term .bar .lights { margin-right: 0; }
+.term .bar .ttl { position: absolute; left: 0; right: 0; text-align: center; pointer-events: none; }
+.term .body { padding: 12px 16px; white-space: pre; font-size: 13px; line-height: 1.7; }
+
+/* e2ee */
+.relay { position: absolute; background: #0d0f12; border: 1px solid #25303d; border-radius: 6px; overflow: hidden; }
+.relay .ph { border-bottom-color: #1a2330; }
+.relay .pt { color: #5b6b7e; font-size: 11.5px; line-height: 1.62; }
+.relay .pt .hot { color: #9fb4cc; }
+.strip { position: absolute; left: 0; right: 0; bottom: 0; height: 34px; display: flex; align-items: center; justify-content: center;
+ gap: 18px; background: var(--tab); border-top: 1px solid #0e0e0e; color: var(--dim); font-size: 11.5px; }
+.strip .lk { color: var(--blue); }
+</style>
+</head>
+<body>
+<div id="app"></div>
+<script>
+const W = 960, H = 600, TOP = 34, GAP = 5, PAD = 5;
+const clamp = (x, a, b) => Math.max(a, Math.min(b, x));
+const esc = s => s.replace(/&/g, '&amp;').replace(/</g, '&lt;').replace(/>/g, '&gt;');
+const typed = (s, t, t0, cps = 24) => esc(s.slice(0, clamp(Math.floor((t - t0) * cps), 0, s.length)));
+const blink = t => Math.floor(t * 2.4) % 2 === 0;
+const cur = (t, on = true) => on ? `<span class="cur${blink(t) ? '' : ' off'}"></span>` : '';
+const stream = (lines, t) => lines.filter(l => t >= l[0]).map(l => l[1]).join('\n');
+const ICON = { claude: '../../engines/claude.png', codex: '../../engines/codex.png', cursor: '../../engines/cursor.png',
+ opencode: '../../engines/opencode.png', terminal: null };
+const svg = {
+ mon: '<svg width="11" height="11" viewBox="0 0 16 16"><rect x="1.5" y="2.5" width="13" height="9" rx="1.5" fill="none" stroke="currentColor" stroke-width="1.6"/><path d="M5 14.2h6" stroke="currentColor" stroke-width="1.6"/></svg>',
+ dir: '<svg width="11" height="11" viewBox="0 0 16 16"><path d="M1.5 4.5v8h13v-7H8L6.5 3.5h-5z" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linejoin="round"/></svg>',
+ br: '<svg width="11" height="11" viewBox="0 0 16 16"><circle cx="4.5" cy="3.5" r="1.8" fill="none" stroke="currentColor" stroke-width="1.5"/><circle cx="4.5" cy="12.5" r="1.8" fill="none" stroke="currentColor" stroke-width="1.5"/><circle cx="11.5" cy="5.5" r="1.8" fill="none" stroke="currentColor" stroke-width="1.5"/><path d="M4.5 5.3v5.4M11.5 7.3c0 3-7 2-7 3.4" fill="none" stroke="currentColor" stroke-width="1.5"/></svg>',
+ bell: '<svg width="14" height="14" viewBox="0 0 16 16"><path d="M8 1.8a4.2 4.2 0 0 0-4.2 4.2v3L2.5 11.3h11L12.2 9V6A4.2 4.2 0 0 0 8 1.8zM6.3 13.2a1.8 1.8 0 0 0 3.4 0" fill="none" stroke="currentColor" stroke-width="1.4" stroke-linejoin="round"/></svg>',
+ lock: '<svg width="11" height="11" viewBox="0 0 16 16"><rect x="3" y="7" width="10" height="7.5" rx="1.5" fill="currentColor"/><path d="M5.2 7V5.2a2.8 2.8 0 0 1 5.6 0V7" fill="none" stroke="currentColor" stroke-width="1.6"/></svg>',
+ store: '<svg width="11" height="11" viewBox="0 0 16 16"><path d="M2.5 6h11l-1 8h-9zM5.5 6V4.5a2.5 2.5 0 0 1 5 0V6" fill="none" stroke="currentColor" stroke-width="1.5" stroke-linejoin="round"/></svg>',
+};
+
+function tabbar({ tab = 'work', badge = 0 } = {}) {
+ return `<div class="tabbar"><div class="lights"><i></i><i></i><i></i></div>
+ <span class="bell">${svg.bell}${badge ? `<span class="badge">${badge}</span>` : ''}</span>
+ <div class="tab">▢ ${tab}</div><span class="plus">+</span>
+ <div class="store">${svg.store} Harness Store</div></div>`;
+}
+
+function pane(p) {
+ const icon = ICON[p.engine] ? `<img src="${ICON[p.engine]}">` : '<span class="dim">&gt;_</span>';
+ const wide = p.w > 420;
+ const meta = `<span>${svg.mon} ${p.machine}</span>` + (wide && p.project ? `<span class="pj">${svg.dir} ${p.project}</span>` : '') +
+ (wide && p.branch ? `<span class="br">${svg.br} ${p.branch}</span>` : '');
+ const style = `left:${p.x}px;top:${p.y}px;width:${p.w}px;height:${p.h}px;${p.opacity != null ? `opacity:${p.opacity};` : ''}`;
+ return `<div class="pane${p.focus ? ' focus' : ''}" style="${style}"><div class="ph">${icon}<b>${p.name}</b>
+ <span class="dim">${p.title}</span><div class="meta">${meta}</div></div><div class="pt">${p.body}</div></div>`;
+}
+
+// Tile rectangles inside the workspace, which ends where the command box starts.
+function area(boxH = 0) { return { x: PAD, y: TOP + PAD, w: W - 2 * PAD, h: H - TOP - 2 * PAD - boxH }; }
+function grid2x2(a) {
+ const w = (a.w - GAP) / 2, h = (a.h - GAP) / 2;
+ return [{ x: a.x, y: a.y, w, h }, { x: a.x + w + GAP, y: a.y, w, h },
+ { x: a.x, y: a.y + h + GAP, w, h }, { x: a.x + w + GAP, y: a.y + h + GAP, w, h }];
+}
+const lerp = (a, b, k) => a + (b - a) * k;
+const ease = k => { k = clamp(k, 0, 1); return k * k * (3 - 2 * k); };
+const mixRect = (r1, r2, k) => ({ x: lerp(r1.x, r2.x, k), y: lerp(r1.y, r2.y, k), w: lerp(r1.w, r2.w, k), h: lerp(r1.h, r2.h, k) });
+
+function hud(t, presses, bottom = 0) {
+ // presses: [time, keys[], label]; each shows for 1.1 s.
+ const p = presses.filter(x => t >= x[0] && t < x[0] + 1.1).pop();
+ if (!p) return '';
+ const k = (t - p[0]);
+ const op = k < 0.08 ? k / 0.08 : k > 0.9 ? (1.1 - k) / 0.2 : 1;
+ return `<div class="hud" style="opacity:${op};bottom:${bottom + 22}px">${p[1].map(x => `<kbd>${x}</kbd>`).join('')}${p[2] ? `<span class="label">${p[2]}</span>` : ''}</div>`;
+}
+
+function boxHeight(kind, rows = 0) {
+ if (kind === 'form') return 186;
+ if (kind === 'picker') return 64 + rows * 23 + 34;
+ return 0;
+}
+function formBox({ title, agent, machine, project, task, t, focusTask = true }) {
+ return `<div class="box" style="height:${boxHeight('form')}px"><div class="head"><span>${title}</span><span></span></div>
+ <div class="field"><span class="k">Agent</span><span>${agent}</span></div>
+ <div class="field"><span class="k">Machine</span><span>${machine}</span></div>
+ <div class="field"><span class="k">Project</span><span>${project}</span></div>
+ <div class="field task"><span class="k">task &gt;</span><span class="v">${task}${cur(t, focusTask)}</span></div>
+ <div class="foot"><span><b>enter</b>start</span><span><b>tab</b>agent · machine · project</span><span><b>⌘.</b>full form</span><span><b>esc</b>close</span></div></div>`;
+}
+function pickerBox({ title, count, rows, query, t, preview, hints }) {
+ const h = boxHeight('picker', Math.max(rows.length, preview ? 5 : 0));
+ return `<div class="box" style="height:${h}px"><div class="head"><span>${title}</span><span>${count || ''}</span></div>
+ <div class="list"><div class="rows">${rows.join('')}</div>${preview ? `<div class="preview">${preview}</div>` : ''}</div>
+ <div class="query">${query}${cur(t)}</div>
+ <div class="foot">${hints}</div></div>`;
+}
+const hint = (k, v) => `<span><b>${k}</b>${v}</span>`;
+
+// ------------------------------------------------------------------ transcripts
+const claudeAuth = t => stream([
+ [0.0, '<span class="dim">&gt;</span> refactor auth to use the session cache'],
+ [0.5, ''],
+ [0.6, '<span class="o">⏺</span> Read(src/auth/middleware.ts)'],
+ [1.4, '<span class="o">⏺</span> Update(src/auth/middleware.ts)'],
+ [1.6, '<span class="dim"> ⎿ +18 −31</span>'],
+ [2.6, '<span class="o">⏺</span> Bash(npm test -- auth)'],
+ [3.7, '<span class="dim"> ⎿ </span><span class="g">42 passed</span>'],
+ [4.5, '<span class="o">⏺</span> Update(src/auth/session.ts)'],
+ [5.0, '<span class="dim"> ⎿ +6 −2</span>'],
+], t) + `\n\n<span class="o">✻ Cooking… (${72 + Math.floor(t)}s · esc to interrupt)</span>`;
+
+const codexSearch = t => stream([
+ [0.0, '<span class="dim">›</span> add full-text search to /products'],
+ [0.4, ''],
+ [0.9, '<span class="dim">•</span> Read prisma/schema.prisma'],
+ [1.7, '<span class="dim">•</span> Planned 3 steps'],
+ [2.4, '<span class="dim">•</span> Wrote migrations/0043_search.sql'],
+ [3.3, '<span class="dim">•</span> Ran npx prisma migrate dev'],
+ [3.6, '<span class="dim"> └ </span><span class="g">✓ migration applied</span>'],
+ [4.7, '<span class="dim">•</span> Edited app/products/page.tsx <span class="g">+64</span> <span class="r">−3</span>'],
+], t) + `\n\n<span class="dim">◦ Working (${31 + Math.floor(t)}s • esc to interrupt)</span>`;
+
+const cursorNav = t => stream([
+ [0.0, 'rebuild the nav for the new sidebar'],
+ [0.4, ''],
+ [0.7, '<span class="g">●</span> Thought 3s'],
+ [1.8, '<span class="g">●</span> Read 3 files, 1 search'],
+ [3.0, '<span class="g">●</span> Edited src/Nav.tsx <span class="g">+41</span> <span class="r">−9</span>'],
+ [4.1, '<span class="g">●</span> Edited src/Sidebar.tsx <span class="g">+12</span> <span class="r">−30</span>'],
+], t) + `\n\n<span class="g">● Working 2m ${String(8 + Math.floor(t)).padStart(2, '0')}s</span>`;
+
+const openEval = t => stream([
+ [0.0, '<span class="y">┃</span> run the eval suite on step-4200'],
+ [0.4, ''],
+ [0.9, '<span class="dim">+ Thought: 117ms</span>'],
+ [1.4, '<span class="dim">→</span> bash python eval.py --ckpt step-4200'],
+ [2.5, '<span class="dim"> mmlu </span>71.8 <span class="g">+1.2</span>'],
+ [3.2, '<span class="dim"> gsm8k </span>84.2 <span class="g">+3.0</span>'],
+ [3.9, '<span class="dim"> humaneval </span>67.1 <span class="g">+2.4</span>'],
+ [4.8, '<span class="dim">→</span> write results/step-4200.json'],
+], t);
+
+// ------------------------------------------------------------------ scenes
+const SCENES = {};
+
+// Every agent: four agents on three machines, all working at once.
+SCENES.agents = { dur: 7.5, draw(t) {
+ const r = grid2x2(area());
+ return tabbar({ tab: 'work' }) +
+ pane({ ...r[0], engine: 'claude', name: 'Claude Code', title: 'auth', machine: 'macbook-pro', project: 'api', branch: 'main', body: claudeAuth(t) }) +
+ pane({ ...r[1], engine: 'codex', name: 'Codex', title: 'search', machine: 'macbook-pro', project: 'shop', branch: 'feat/search', body: codexSearch(t) }) +
+ pane({ ...r[2], engine: 'cursor', name: 'Cursor', title: 'nav', machine: 'office-desktop', project: 'mobile', branch: 'main', body: cursorNav(t) }) +
+ pane({ ...r[3], engine: 'opencode', name: 'OpenCode', title: 'eval', machine: 'gpu-box', project: 'llm', branch: 'main', body: openEval(t) });
+} };
+
+// Every machine: set up a new box in four commands, link it, start an agent there.
+SCENES.machines = { dur: 12.5, draw(t) {
+ let html = tabbar({ tab: 'work' });
+ const base = [
+ { engine: 'claude', name: 'Claude Code', title: 'auth', machine: 'macbook-pro', project: 'api', branch: 'main', body: claudeAuth(t + 6) },
+ { engine: 'codex', name: 'Codex', title: 'search', machine: 'macbook-pro', project: 'shop', branch: 'feat/search', body: codexSearch(t + 6) },
+ ];
+ // Box states over time.
+ let box = '', boxH = 0;
+ if (t >= 4.6 && t < 7.4) {
+ boxH = boxHeight('picker', 4);
+ const linking = t >= 5.6, linked = t >= 6.7;
+ const rows = [
+ `<div class="row"><span class="t">macbook-pro</span><span class="m">this Mac</span><span class="g">● online</span></div>`,
+ `<div class="row"><span class="t">home-server</span><span class="m">linked</span><span class="g">● online</span></div>`,
+ `<div class="row"><span class="t">office-desktop</span><span class="m">linked</span><span class="g">● online</span></div>`,
+ `<div class="row sel"><span class="t">gpu-box</span>${linked ? '<span class="g">✓ linked · AA24·B561·0B81·EF21</span>' : '<span class="m">Link this machine · Enter its remote password</span>'}<span class="ret">↵</span></div>`,
+ ];
+ const query = linking && !linked ? `<span class="dim">password </span>${'•'.repeat(clamp(Math.floor((t - 5.7) * 14), 0, 12))}` : linked ? '' : 'gpu';
+ box = pickerBox({ title: 'Machines', count: '4 machines', rows, query, t,
+ hints: hint('↑↓', 'select') + hint('enter', 'link') + hint('esc', 'close') });
+ } else if (t >= 7.6 && t < 10.0) {
+ boxH = boxHeight('form');
+ box = formBox({ title: 'New Harness · work', agent: '<img src="../../engines/claude.png" style="width:12px;height:12px;vertical-align:-1px;border-radius:2px"> Claude Code',
+ machine: '<span class="w">gpu-box</span>', project: '~/tokenizer', task: typed('train the tokenizer on the new corpus', t, 7.9, 26), t });
+ }
+ const a = area(boxH);
+ const split = t >= 10.0;
+ const k = ease((t - 10.0) / 0.35);
+ const w3 = (a.w - 2 * GAP) / 3;
+ const two = [{ x: a.x, y: a.y, w: (a.w - GAP) / 2, h: a.h }, { x: a.x + (a.w + GAP) / 2, y: a.y, w: (a.w - GAP) / 2, h: a.h }];
+ const three = [{ x: a.x, y: a.y, w: w3, h: a.h }, { x: a.x + w3 + GAP, y: a.y, w: w3, h: a.h }, { x: a.x + 2 * (w3 + GAP), y: a.y, w: w3, h: a.h }];
+ base.forEach((p, i) => { html += pane({ ...p, ...(split ? mixRect(two[i], three[i], k) : two[i]) }); });
+ if (split) {
+ const s = t - 10.1;
+ const enter = { x: a.x + a.w + GAP, y: a.y, w: w3, h: a.h };
+ html += pane({ ...mixRect(enter, three[2], k), focus: true, engine: 'claude', name: 'Claude Code', title: 'tokenizer', machine: 'gpu-box', project: 'tokenizer', branch: 'main',
+ body: stream([
+ [0.0, '<span class="dim">&gt;</span> train the tokenizer on the new corpus'],
+ [0.1, ''],
+ [0.4, '<span class="o">⏺</span> Bash(nvidia-smi -L)'],
+ [0.9, '<span class="dim"> ⎿ </span>8 × NVIDIA H100 80GB'],
+ [1.3, '<span class="o">⏺</span> Bash(python train.py --vocab 64k)'],
+ ], s) + (s > 1.6 ? `\n\n<span class="o">✻ Training… epoch 1/3</span>` : '') });
+ }
+ html += box;
+ // The new machine's own terminal, shown first.
+ if (t < 4.4) {
+ const op = t > 4.0 ? 1 - (t - 4.0) / 0.4 : 1;
+ const cmd = (s, t0) => typed(s, t, t0, 34);
+ const lines = [];
+ const steps = [
+ [0.2, 'curl -fsSL harness.autonomous.ai/cli/install.sh | bash', 1.8, '<span class="g">✓</span> harness installed'],
+ [2.0, 'harness login', 2.5, '<span class="g">✓</span> signed in'],
+ [2.7, 'harness remote-password set', 3.3, '<span class="g">✓</span> password saved'],
+ [3.4, 'harness start', 3.8, '<span class="g">✓</span> gpu-box is online'],
+ ];
+ for (const [t0, c, t1, out] of steps) {
+ if (t < t0) break;
+ lines.push(`<span class="g">ubuntu@gpu-box</span><span class="dim">:~$</span> ${cmd(c, t0)}${t < t1 && (steps.find(s => s[0] > t0)?.[0] ?? 9) > t ? cur(t) : ''}`);
+ if (t >= t1) lines.push(out);
+ }
+ html += `<div class="dimmer" style="opacity:${op}"></div><div class="term" style="opacity:${op}"><div class="bar"><div class="lights"><i></i><i></i><i></i></div><span class="ttl">ubuntu@gpu-box: ~</span></div><div class="body">${lines.join('\n')}</div></div>`;
+ }
+ html += hud(t, [[6.6, ['⏎'], 'link'], [7.6, ['⌘', 'N'], 'new harness'], [9.8, ['⏎'], 'start']], boxH);
+ return html;
+} };
+
+// Keyboard first: open any session, zoom, answer the agent waiting on you, split.
+SCENES.keyboard = { dur: 13, draw(t) {
+ const sessions = {
+ auth: { engine: 'claude', name: 'Claude Code', title: 'auth', machine: 'macbook-pro', project: 'api', branch: 'main' },
+ search: { engine: 'codex', name: 'Codex', title: 'search', machine: 'macbook-pro', project: 'shop', branch: 'feat/search' },
+ tok: { engine: 'claude', name: 'Claude Code', title: 'tokenizer', machine: 'gpu-box', project: 'tokenizer', branch: 'main' },
+ bill: { engine: 'codex', name: 'Codex', title: 'billing', machine: 'home-server', project: 'billing', branch: 'stripe-v2' },
+ };
+ const asking = t >= 5.2 && t < 7.4;
+ const answered = t >= 7.4;
+ const authBody = stream([
+ [0, '<span class="dim">&gt;</span> refactor auth to use the session cache'], [0, ''],
+ [0, '<span class="o">⏺</span> Update(src/auth/middleware.ts)'], [0, '<span class="dim"> ⎿ +18 −31</span>'],
+ [0, '<span class="o">⏺</span> Bash(npm test -- auth)'], [0, '<span class="dim"> ⎿ </span><span class="g">42 passed</span>'],
+ ], t) + (asking ? '\n\n<div class="ask"><span class="w">Edit src/auth/session.ts?</span>\n<span class="b">❯ 1. Yes</span>\n 2. Yes, and don\'t ask again\n 3. No</div>' :
+ answered ? '\n<span class="o">⏺</span> Update(src/auth/session.ts)\n<span class="dim"> ⎿ +6 −2</span>\n\n<span class="o">✻ Cooking… (1m 31s)</span>' : '\n\n<span class="dim">✻ Thinking…</span>');
+ const bodies = {
+ auth: authBody,
+ search: codexSearch(t + 6),
+ tok: stream([[0, '<span class="dim">&gt;</span> train the tokenizer on the new corpus'], [0, ''], [0, '<span class="o">⏺</span> Bash(python train.py --vocab 64k)'], [0, '<span class="dim"> ⎿ epoch 2/3 · loss 1.84</span>']], t) + '\n\n<span class="o">✻ Training…</span>',
+ bill: stream([[0, '<span class="dim">›</span> migrate billing to stripe v2'], [0, ''], [0, '<span class="dim">•</span> Read billing/charge.ts'], [0, '<span class="dim">•</span> Planned 4 steps'], [0, '<span class="dim">•</span> Edited billing/charge.ts <span class="g">+88</span> <span class="r">−52</span>'], [0, '<span class="dim">•</span> Ran npm test billing'], [0, '<span class="dim"> └ </span><span class="g">✓ 31 passed</span>']], t) + '\n\n<span class="dim">◦ Working (2m 04s)</span>',
+ roll: '',
+ };
+ // Picker (⌘O) 0.4 – 2.6
+ let box = '', boxH = 0;
+ if (t >= 0.4 && t < 2.6) {
+ const q = typed('bill', t, 1.0, 9);
+ const all = [
+ ['auth', 'refactor auth', 'Claude', 'macbook-pro', 'api'],
+ ['search', 'full-text search', 'Codex', 'macbook-pro', 'shop'],
+ ['tok', 'train tokenizer', 'Claude', 'gpu-box', 'tokenizer'],
+ ['bill', 'migrate billing', 'Codex', 'home-server', 'billing'],
+ ['nav', 'rebuild the nav', 'Cursor', 'office-desktop', 'mobile'],
+ ];
+ const shown = q ? all.filter(r => (r[1] + r[4]).includes(q.replace(/&[a-z]+;/g, ''))) : all;
+ const sel = shown.length - 1;
+ const rows = shown.map((r, i) => `<div class="row${i === sel ? ' sel' : ''}"><span class="t">${r[1]}</span><span class="e">&gt;_ ${r[2]}</span><span class="m">${svg.mon} ${r[3]}</span><span class="pj">${svg.dir} ${r[4]}</span>${i === sel ? '<span class="ret">↵</span>' : ''}</div>`);
+ const s = shown[sel];
+ const preview = s ? `<span class="t">${s[1]}</span>\n<span class="g">Working</span> · &gt;_ ${s[2]} · ${s[3]}\n\n${s[0] === 'bill' ? '• Edited billing/charge.ts +88 −52\n• Ran npm test billing\n └ ✓ 31 passed' : '• …'}` : '';
+ boxH = boxHeight('picker', 5);
+ box = pickerBox({ title: 'Open Harness · work', count: `${shown.length}/5`, rows, query: q, t, preview,
+ hints: hint('↑↓', 'select') + hint('enter', 'focus pane') + hint('@', 'machines') + hint('#', 'projects') + hint('esc', 'close') });
+ }
+ // Split below (⌘D) 9.0 – 11.2
+ if (t >= 9.0 && t < 11.2) {
+ boxH = boxHeight('form');
+ box = formBox({ title: 'New Pane Below · work', agent: '<img src="../../engines/codex.png" style="width:12px;height:12px;vertical-align:-1px;border-radius:2px"> Codex', machine: 'home-server', project: 'billing', task: typed('write the rollback for the migration', t, 9.4, 24), t });
+ }
+ const a = area(boxH);
+ const g = grid2x2(a);
+ let focus = 'search';
+ if (t >= 2.6) focus = 'bill';
+ if (t >= 5.9) focus = 'auth';
+ if (t >= 11.2) focus = 'roll';
+ // Zoom the billing pane 3.4 – 4.6
+ const zoomIn = ease((t - 3.4) / 0.3), zoomOut = ease((t - 4.6) / 0.3);
+ const zk = t < 4.6 ? zoomIn : 1 - zoomOut;
+ const order = ['auth', 'search', 'tok', 'bill'];
+ let rects = order.map((id, i) => g[i]);
+ // After the split, the billing column shares its height with the rollback pane.
+ const splitK = ease((t - 11.2) / 0.35);
+ if (t >= 11.2) {
+ const b = g[3], h2 = (b.h - GAP) / 2;
+ rects[3] = mixRect(b, { x: b.x, y: b.y, w: b.w, h: h2 }, splitK);
+ }
+ let html = tabbar({ tab: 'work', badge: asking ? 1 : 0 });
+ order.forEach((id, i) => {
+ if (id === 'bill') return;
+ html += pane({ ...sessions[id], ...rects[i], focus: focus === id, body: bodies[id], opacity: zk > 0 ? 1 - zk : 1 });
+ });
+ html += pane({ ...sessions.bill, ...mixRect(rects[3], a, zk), focus: focus === 'bill', body: bodies.bill });
+ if (t >= 11.2) {
+ const b = g[3], h2 = (b.h - GAP) / 2;
+ html += pane({ ...mixRect({ x: b.x, y: b.y + b.h, w: b.w, h: 0 }, { x: b.x, y: b.y + h2 + GAP, w: b.w, h: h2 }, splitK), focus: true, engine: 'codex', name: 'Codex', title: 'rollback', machine: 'home-server', project: 'billing', branch: 'stripe-v2',
+ body: stream([[0, '<span class="dim">›</span> write the rollback for the migration'], [0.3, ''], [0.5, '<span class="dim">•</span> Read migrations/0012_stripe_v2.sql'], [1.1, '<span class="dim">•</span> Planned 2 steps']], t - 11.3) });
+ }
+ html += box;
+ html += hud(t, [[0.3, ['⌘', 'O'], 'open anything'], [2.5, ['⏎'], ''], [3.3, ['⌘', '⏎'], 'zoom'], [4.5, ['⌘', '⏎'], ''],
+ [5.8, ['⇧', '⌘', 'I'], 'agent needs you'], [7.3, ['1'], 'yes'], [8.9, ['⌘', 'D'], 'split down'], [11.1, ['⏎'], '']], boxH);
+ return html;
+} };
+
+// End-to-end encrypted: what you see, and what the relay carries.
+function hex(seed, n) {
+ let x = seed * 2654435761 >>> 0, s = '';
+ for (let i = 0; i < n; i++) { x ^= x << 13; x ^= x >>> 17; x ^= x << 5; x >>>= 0; s += (x & 0xffff).toString(16).padStart(4, '0') + ' '; }
+ return s.trim();
+}
+SCENES.e2ee = { dur: 8, draw(t) {
+ const a = { x: PAD, y: TOP + PAD, w: W - 2 * PAD, h: H - TOP - 2 * PAD - 34 };
+ const w = (a.w - GAP) / 2;
+ const src = [
+ [0.0, '<span class="dim">&gt;</span> rotate the stripe key in .env and redeploy'], [0.3, ''],
+ [0.8, '<span class="o">⏺</span> Read(.env.production)'],
+ [1.6, '<span class="o">⏺</span> Update(.env.production)'],
+ [2.0, '<span class="dim"> ⎿ STRIPE_KEY=sk_live_••••••••4f2a</span>'],
+ [3.0, '<span class="o">⏺</span> Bash(fly deploy --app shop)'],
+ [4.2, '<span class="dim"> ⎿ </span><span class="g">✓ v142 deployed</span>'],
+ [5.2, '<span class="o">⏺</span> Bash(curl -s shop.dev/health)'],
+ [5.9, '<span class="dim"> ⎿ </span><span class="g">ok</span>'],
+ ];
+ const left = stream(src, t) + (t < 6.3 ? `\n\n<span class="o">✻ Working… (${12 + Math.floor(t)}s)</span>` : '');
+ // Every screen update becomes one sealed frame on the wire.
+ const frames = [];
+ let n = 0;
+ for (let k = 0; k * 0.18 <= t; k++) {
+ const tt = k * 0.18;
+ const busy = src.some(l => l[0] <= tt && tt - l[0] < 0.5) || k % 3 === 0;
+ if (!busy) continue;
+ n++;
+ frames.push(`<span class="dim2">${String(1840 + n).padStart(5, ' ')}</span> ${hex(k + 7, 7)}`);
+ }
+ const shown = frames.slice(-26).map((l, i, arr) => i === arr.length - 1 ? `<span class="hot">${l}</span>` : l).join('\n');
+ return tabbar({ tab: 'deploy' }) +
+ pane({ x: a.x, y: a.y, w, h: a.h, focus: true, engine: 'claude', name: 'Claude Code', title: 'deploy', machine: 'macbook-pro', project: 'shop', branch: 'main', body: left }) +
+ `<div class="relay" style="left:${a.x + w + GAP}px;top:${a.y}px;width:${w}px;height:${a.h}px"><div class="ph"><span class="b">${svg.lock}</span><b>What the relay sees</b><span class="dim">harness relay</span><div class="meta"><span>ciphertext only</span></div></div><div class="pt">${shown}</div></div>` +
+ `<div class="strip"><span class="lk">${svg.lock}</span><span>ChaCha20-Poly1305</span><span>X25519</span><span>pinned Ed25519 keys</span><span>always on, no switch</span></div>`;
+} };
+
+function draw(scene, t) { document.getElementById('app').innerHTML = SCENES[scene].draw(t); }
+window.SCENES = SCENES; window.draw = draw;
+if (location.hash) {
+ const name = location.hash.slice(1), t0 = performance.now();
+ const tick = () => { draw(name, ((performance.now() - t0) / 1000) % SCENES[name].dur); requestAnimationFrame(tick); };
+ tick();
+}
+</script>
+</body>
+</html>
diff --git a/.github/assets/readme/site.mjs b/.github/assets/readme/site.mjs
new file mode 100644
index 000000000..47cc792bd
--- /dev/null
+++ b/.github/assets/readme/site.mjs
@@ -0,0 +1,69 @@
+#!/usr/bin/env node
+// GIFs made from autonomous.ai pages:
+// device.gif the Harness device video on autonomous.ai/harness-device
+// connect/direct.gif the three connection animations on autonomous.ai/harness-app
+// connect/cloudflare.gif
+// connect/relay.gif
+//
+// node .github/assets/readme/site.mjs
+//
+// Needs Google Chrome, FFmpeg, curl, and playwright-core from store/tools/experience-tests.
+import { execFileSync } from 'node:child_process';
+import { mkdirSync, readdirSync, rmSync, statSync } from 'node:fs';
+import { dirname, join, resolve } from 'node:path';
+import { fileURLToPath, pathToFileURL } from 'node:url';
+
+const here = dirname(fileURLToPath(import.meta.url));
+const root = resolve(here, '../../..');
+const tmp = join(here, '.cache', 'site');
+mkdirSync(join(here, 'connect'), { recursive: true });
+mkdirSync(tmp, { recursive: true });
+const ff = (...a) => execFileSync('ffmpeg', ['-v', 'error', '-y', ...a]);
+const gif = (src, out, vf, extra = []) => {
+ ff(...extra, '-i', src, '-filter_complex',
+ `${vf},split[a][b];[a]palettegen=max_colors=192:stats_mode=diff[p];[b][p]paletteuse=dither=none:diff_mode=rectangle`,
+ '-loop', '0', out);
+ console.log(`${out.slice(here.length + 1)} ${(statSync(out).size / 1048576).toFixed(1)} MiB`);
+};
+
+// The device: tap, speak a task, watch it ship, read the summary.
+const video = join(tmp, 'device.mp4');
+execFileSync('curl', ['-fsSL', '-o', video, 'https://cdn.autonomous.ai/production/ecm/260918/gallery.mp4']);
+gif(video, join(here, 'device.gif'), 'fps=10,scale=960:-2:flags=lanczos');
+
+// The connection animations run in the page itself, so record them in Chrome.
+const { chromium } = await import(pathToFileURL(join(root, 'store/tools/experience-tests/node_modules/playwright-core/index.mjs')).href);
+const browser = await chromium.launch({ channel: 'chrome' });
+const tabs = [['Machine to machine', 'direct'], ['Cloudflare', 'cloudflare'], ['Our relay', 'relay']];
+for (const [tab, name] of tabs) {
+ const dir = join(tmp, name);
+ rmSync(dir, { recursive: true, force: true });
+ const ctx = await browser.newContext({ viewport: { width: 1440, height: 900 }, recordVideo: { dir, size: { width: 1440, height: 900 } } });
+ const page = await ctx.newPage();
+ await page.goto('https://www.autonomous.ai/harness-app', { waitUntil: 'networkidle' });
+ const button = page.getByRole('button', { name: tab, exact: true }).first();
+ await button.scrollIntoViewIfNeeded();
+ await button.click();
+ // Hide the site header and banners, then centre the animation beside the tab row.
+ const box = await page.evaluate((tab) => {
+ for (const el of document.querySelectorAll('body *')) {
+ const pos = getComputedStyle(el).position;
+ if (pos === 'fixed' || pos === 'sticky') el.style.visibility = 'hidden';
+ }
+ const b = [...document.querySelectorAll('button')].find(x => x.innerText.trim() === tab);
+ const below = b.getBoundingClientRect().bottom;
+ const el = [...document.querySelectorAll('div')].map(d => ({ d, r: d.getBoundingClientRect() }))
+ .filter(x => x.r.top > below - 5 && x.r.top < below + 200 && x.r.left > 400 && x.r.width > 600 && x.r.height > 300)
+ .sort((a, b) => b.r.width * b.r.height - a.r.width * a.r.height)[0].d;
+ el.scrollIntoView({ block: 'center' });
+ const r = el.getBoundingClientRect();
+ return { x: Math.round(r.x), y: Math.round(r.y), w: Math.round(r.width), h: Math.round(r.height) };
+ }, tab);
+ await page.waitForTimeout(17000);
+ await ctx.close();
+ const webm = join(dir, readdirSync(dir)[0]);
+ // The loop is steady after the page settles; keep eight seconds of it.
+ gif(webm, join(here, 'connect', `${name}.gif`),
+ `crop=${box.w}:${box.h}:${box.x}:${box.y},fps=12,scale=800:-2:flags=lanczos`, ['-ss', '8', '-t', '8']);
+}
+await browser.close();
diff --git a/README.md b/README.md
index 28e3c723a..889921f0b 100644
--- a/README.md
+++ b/README.md
@@ -1,250 +1,237 @@
-> “World-class entrepreneurs are polymaths.”
->
-> — [Peter Thiel](https://www.youtube.com/watch?v=h10kXgTdhNU&t=811s)
+<p align="center">
+ <img src="desktop/assets/app_icon.png" width="88" alt="">
+</p>
-# Follow your curiosity. Build across disciplines.
+<h1 align="center">OpenHarness</h1>
-OpenHarness is open-source software and hardware for building with coding agents. Use Claude Code,
-Codex, and others to make software, design parts and circuit boards, explore data, and create videos,
-with your agents, domain tools, and live viewers in one place and sessions that persist across every
-machine you own.
+<p align="center">
+ <b>The terminal for coding agents.</b><br>
+ Every agent. Every machine. One fast, keyboard-first window.
+</p>
-The old rule said ten thousand hours to a craft. That was the tax on curiosity, and most of us could
-only afford to pay it once, so we specialized and called the rest hobbies.
+<p align="center">
+ <a href="https://harness.autonomous.ai/desktop"><b>Download</b></a> ·
+ <a href="#how-it-works">How it works</a> ·
+ <a href="#beyond-code">Beyond code</a> ·
+ <a href="#domain-specific-harnesses-dsh">Harnesses</a> ·
+ <a href="#harness-device">Device</a>
+</p>
-Coding agents become the specialists, given the tools of a craft and a way to see what they made:
-the geometry that becomes a part, the netlist that becomes a circuit board, the script that becomes
-a film. You bring the idea, the taste, and the judgment about what is worth making.
+### Every agent, side by side
-Use the desktop app on its own, or add the optional **Harness device** to follow your agents, answer
-their questions, and speak new tasks from your desk. The app, firmware, schematics, PCB layouts, and
-enclosure CAD files are all open source.
+Claude Code, Codex, Cursor, OpenCode, Devin, Amp, Copilot and seven more.
-[Who it's for](#for-polymaths-in-the-making) · [What you can make](#domain-specific-harnesses-dsh) · [Get hands-on](docs/hands-on.md) · [Run it](#run-it) ·
-[Harness device](#harness-device) · [Architecture](docs/architecture.md) · [Contribute](#contributing)
+<p align="center"><img src=".github/assets/readme/agents.gif" width="960" alt="Four agents working at once in one window: Claude Code and Codex on a MacBook, Cursor on an office desktop, OpenCode on a GPU box. Each pane shows its machine, project and branch."></p>
-<p align="center">
- <a href=".github/assets/store/workshop-overview.svg"><img src=".github/assets/store/workshop-overview.svg" width="1280" alt="Start with Coding: 14 agents including Claude Code, Codex, and Cursor. Then explore Design, Engineering, Media, Music, Productivity, Science and Data, Simulation, Games, Research, and Local AI."></a>
-</p>
+### Every machine, side by side
-## For polymaths in the making.
+Your laptop, home server and GPU box in one window. Link each one with a password.
+No SSH keys. No Tailscale. No port forwarding.
-Built for the curious engineer who wants to build beyond software. Get your hands on more of what
-you are making, from the physical product to the analysis and the launch video. Makers and creators
-with the same curiosity are welcome. You get there by directing a specialist in each craft and
-judging what comes back, not by mastering them all first.
+<p align="center"><img src=".github/assets/readme/machines.gif" width="960" alt="A new GPU box runs four setup commands and comes online. The app links it with its password, then starts Claude Code there."></p>
-Monday, a feature. Tuesday, the customer data. Wednesday, an enclosure for the prototype. Thursday,
-the launch video. You already know how to build with code, and Harness brings that way of working to
-CAD, circuit boards, games, videos, and documents, with agents and tools for each craft.
+### Keyboard first
-Your ideas can take you into unfamiliar crafts: an app, a physical product, a game, a film. This is
-a place to follow them. Give each job an agent with the right tools, inspect what it makes, and steer
-the next iteration. The code and project files are there to read, change, version, and build on.
+⌘O finds any session on any machine. ⇧⌘I jumps to the agent waiting on you. ⌘D splits. Every key remaps.
-**From handoff to hands-on.** You bring intent and judgment; your agents write and run code. Each
-harness supplies the tools and feedback for a different craft. You can get your hands on more of
-the product and the work of bringing it to customers.
+<p align="center"><img src=".github/assets/readme/keyboard.gif" width="960" alt="Keyboard only: open a session by typing a few letters, zoom it, jump to the agent asking a question and answer it, then split a new pane below."></p>
-Start with what you know and learn the next craft through the things you build. The
-[ideal-user guide](docs/ideal-users.md) records who we're building for and how the app, community,
-and device serve them.
+### End-to-end encrypted
-## Domain-specific harnesses (DSH)
+Code, keys and keystrokes are sealed on your machine. The relay forwards bytes it can't read.
-**Coding agents can build far more than software.** Code is the common medium: geometry scripts make
-parts, animation code makes videos, and analysis code turns data into charts and findings. That is
-what connects the dots between crafts, and a harness gives the agent the tools and feedback to work
-in each one.
+<p align="center"><img src=".github/assets/readme/e2ee.gif" width="960" alt="On the left, an agent rotates a secret and redeploys. On the right, the same session as the relay sees it: numbered frames of ciphertext."></p>
-<img width="1000" height="563" alt="harness" src="https://github.com/user-attachments/assets/25c63335-279a-4f9a-b702-2ffaec2b096f" />
+### Fast and light
+A native app, not Electron. Close it and your agents keep working.
-A **domain-specific harness** turns a coding agent into a specialist. It brings the domain's
-instructions and skills, a pinned toolchain, a project template, a verdict the app can read, and a
-**live viewer** for what the agent makes. You chat on one side; the board, the part, the robot or the
-game takes shape on the other, and stays interactive after the agent is done.
+| Action | Median | p95 |
+|---|---:|---:|
+| ⌘N new harness | 13.5 ms | 14.6 ms |
+| ⌘O open anything | 15.7 ms | 18.4 ms |
+| ⌘P every command | 14.7 ms | 15.2 ms |
+| ⌘F find in a terminal | 12.5 ms | 18.3 ms |
+| Focus a pane | 11.4 ms | 17.2 ms |
+| Zoom a pane | 14.8 ms | 17.2 ms |
+| Next tab | 17.8 ms | 24.2 ms |
-A DSH is a folder with a `harness.json`. The agent does the reasoning; the harness brings the tools
-and the view. Adding a domain never needs a change to the app or the daemon.
+Key dispatch to finished frame. Release build, M2 Max, 16 live terminals. A window in the background
+runs zero timers. [How we measure](https://github.com/autonomous-ai/openharness/blob/codex/perf-integration-checkpoint/docs/performance/2026-09-22-desktop-latency.md).
-**Try the part only you can do.** Change a robot's world, shape a lamp, rewind a jump, perform a
-track, turn a molecular bond, review a draft, sharpen a question, or compare a circuit's traces.
-The [hands-on field guide](docs/hands-on.md) has eight native demos, starting prompts and ways to
-keep your decisions. Open [the interactive guide](docs/hands-on.html) from this checkout in a browser
-to pick a craft and watch the local recordings.
-For a live preview from this checkout, [run a native starter](docs/try-hands-on.md) for physics,
-music or circuits using your installed harness assets.
+### Built the way developers work
-<!-- store-catalog:start -->
-### Coding and beyond
+- **Real terminals.** Every agent runs in its own tmux pane. Scrollback, colors and keys just work.
+- **Your CLIs, as they are.** Harness never wraps an agent. It reads transcripts and uses the vendor's own hooks.
+- **A worktree per harness.** Start an agent on its own branch. Your working copy stays clean.
+- **Remap every key.** One JSONC file, reloaded on save. Chords up to four strokes.
+- **Local models.** Run open-weight models on your own machines with Grid, Ollama, MLX-LM and vLLM.
+- **Bytes, not pixels.** Remote terminals stream text peer to peer. No remote desktop.
+- **Open source, all of it.** App, CLI, daemon, relay and device, in this repo.
-Start with a coding agent you already use. Explore 49 domain-specific harnesses when your
-next idea takes you further.
+## How it works
-| Category | Agents and harnesses |
-|---|---|
-| **Coding** | [Claude Code, Codex, Cursor, OpenCode, Pi, Hermes, Command Code, Devin, Muse Code, Amp, Antigravity, GitHub Copilot, Grok Build, Kilo Code](docs/engines.md), [Harness Monitor](store/agents/harness-monitor/), [Machine Monitor](store/agents/machine-monitor/) |
-| Design | [Autonomous Workshop](store/agents/autonomous-workshop/), [Blender](store/agents/blender/), [Bonsai MCP](store/agents/bonsai-mcp/), [Creative Direction](store/agents/creative-direction/), [Excalidraw](store/agents/excalidraw/), [FreeCAD](store/agents/freecad/), [Generative Art](store/agents/generative-art/), [OpenSCAD](store/agents/openscad/), [text-to-cad](store/agents/text-to-cad/) |
-| Engineering | [Autonomous Circuit](store/agents/autonomous-circuit/), [CircuitJS](store/agents/circuitjs/), [Home Assistant](store/agents/home-assistant/), [KiCad](store/agents/kicad/), [Orca Slicer](store/agents/orca-slicer/), [Yosys](store/agents/yosys/) |
-| Media | [Comfy MCP](store/agents/comfy-mcp/), [Manim](store/agents/manim/), [OpenMontage](store/agents/openmontage/), [Remotion](store/agents/remotion/) |
-| Music | [Ableton AI](store/agents/ableton-ai/), [JUCE Agent Toolkit](store/agents/juce-agent-toolkit/), [Music Studio](store/agents/music-studio/), [Score](store/agents/score/), [Strudel](store/agents/strudel/) |
-| Productivity | [Jev Sheets](store/agents/jev-sheets/), [Marp](store/agents/marp/), [Typst](store/agents/typst/) |
-| Science & Data | [autoresearch-mlx](store/agents/autoresearch-mlx/), [Data Studio](store/agents/data-studio/), [Lab Bench](store/agents/lab-bench/), [marimo](store/agents/marimo/), [RDKit](store/agents/rdkit/) |
-| Simulation | [DimOS](store/agents/dimos/), [Drone Pilot](store/agents/drone-pilot/), [Foam-Agent](store/agents/foam-agent/), [MuJoCo](store/agents/mujoco/), [SimSkill](store/agents/simskill/) |
-| Games | [Game Master](store/agents/game-master/), [Godogen](store/agents/godogen/), [Phaser](store/agents/phaser/), [Voxel Worlds](store/agents/voxel-worlds/) |
-| Research | [Jev Browser](store/agents/jev-browser/), [Roundtable](store/agents/roundtable/) |
-| Local AI | [Grid](store/agents/autonomous-grid/), [MLX-LM](store/agents/mlx-lm/), [Ollama](store/agents/ollama/), [vLLM](store/agents/vllm/) |
+One daemon per machine runs your agents in tmux. It dials out, so no machine opens a port.
-These are the 49 harnesses currently listed in the Store catalog. They combine upstream
-open-source tools and original workflows, with instructions, setup, checks, and live views for each craft.
+```mermaid
+flowchart LR
+ subgraph laptop["Your laptop"]
+ app["Harness app"] -- loopback --> d1["daemon"]
+ d1 --> a1["tmux · Claude Code · Codex"]
+ end
+ subgraph box["Your GPU box"]
+ d2["daemon"] --> a2["tmux · OpenCode · Hermes"]
+ end
+ device["Harness device"] -- USB --> d1
+ d1 <-. "direct WebRTC" .-> d2
+ d1 <-- "ciphertext" --> relay["Harness relay · no keys"]
+ relay <-- "ciphertext" --> d2
+```
-The 10 [shared viewers](store/viewers/) cover CAD, 3D models, documents, games, film, video,
-MuJoCo, web pages, isolated web previews, and studios. Viewer packages install alongside the
-harnesses that need them. Experimental packages marked unlisted are not included above.
-<!-- store-catalog:end -->
+Every path is sealed end to end: ChaCha20-Poly1305, X25519 session keys, pinned Ed25519 identities.
+Harness picks the best path on its own.
-**The harness we'd love to see next is the one for your craft.** Bring an open-source tool you use,
-a workflow you know well, or your own company's toolchain. A harness can live in this repository or
-in yours.
+**Direct.** A WebRTC channel between your machines. No server in the path.
-<!-- store-showcase:start -->
-<p align="center">
- <a href=".github/assets/store/showcase.gif"><img src=".github/assets/store/showcase.gif" width="1280" alt="Six real harness outputs, shown one at a time with their harness name and complete prompt: Autonomous Circuit, text-to-cad, MuJoCo, Blender, Godogen, and Manim."></a>
-</p>
+<p align="center"><img src=".github/assets/readme/connect/direct.gif" width="800" alt="The Harness window connects straight to home-server, office-desktop, cloud-server and mac-mini over WebRTC, with no server in the path."></p>
+
+**Through Cloudflare.** When a firewall blocks the direct path, the same encrypted channel runs over
+Cloudflare's TURN network. Harness keeps trying for a direct path.
+
+<p align="center"><img src=".github/assets/readme/connect/cloudflare.gif" width="800" alt="The encrypted WebRTC channel passes through the nearest Cloudflare edge when a firewall blocks the direct path."></p>
+
+**Through our relay.** A fallback while WebRTC negotiates. The relay holds no keys and forwards ciphertext.
+
+<p align="center"><img src=".github/assets/readme/connect/relay.gif" width="800" alt="The Harness relay keeps the session reachable over its WebSocket while WebRTC negotiates. It holds no keys."></p>
+
+The [architecture guide](docs/architecture.md) has the details.
-Six real outputs, one at a time. Each slide includes the harness and the original prompt.
-[Still preview](.github/assets/store/showcase-poster.png) · Individual images and prompts below.
+<a id="run-it"></a>
+## Get started
+
+**[Download the app](https://harness.autonomous.ai/desktop)** for macOS or Linux.
+
+Add a machine. Run this on it, then **Machines → Link Machine** in the app:
+
+```bash
+curl -fsSL https://harness.autonomous.ai/cli/install.sh | bash
+harness login
+harness remote-password set
+harness start
+```
<details>
-<summary>Read the prompts and open individual images</summary>
+<summary><b>Build from source</b></summary>
-**[Autonomous Circuit](store/showcase/autonomous-circuit/six-key-macropad.jpg)** · [Open harness](store/agents/autonomous-circuit/)
+Needs Node.js 20+, tmux, Xcode and Flutter 3.47+ / Dart 3.13+:
-> Design a six-key USB macropad. Start with the schematic.
+```bash
+git clone https://github.com/autonomous-ai/openharness.git
+cd openharness
+(cd cli && npm ci)
+make install-cli
+cd desktop
+flutter config --enable-swift-package-manager
+flutter pub get
+flutter run -d macos
+```
-**[text-to-cad](store/showcase/text-to-cad/planetary-gear-set.jpg)** · [Open harness](store/agents/text-to-cad/)
+`make install-cli` installs this checkout's CLI and restarts the local daemon. See the
+[development guide](docs/development.md).
-> Design a 3D-printable planetary gear set: a 12-tooth sun, three 18-tooth planets and a 48-tooth ring gear with mounting lugs, module 1.5 and 8 mm thick, plus a carrier on steel pins. Give each part its own colour.
+</details>
-**[MuJoCo](store/showcase/mujoco/g1-humanoid-hello.jpg)** · [Open harness](store/agents/mujoco/)
+## Beyond code
-> Make the Unitree G1 humanoid say hello: stand, raise its right hand and wave three times, then lower it and take a small bow. Record it.
+> “World-class entrepreneurs are polymaths.” — [Peter Thiel](https://www.youtube.com/watch?v=h10kXgTdhNU&t=811s)
-**[Blender](store/showcase/blender/cozy-reading-nook.jpg)** · [Open harness](store/agents/blender/)
+**Coding agents can build far more than software.** Give one a harness and it works with the real
+tools of a craft. You steer in a live viewer. Every clip below is a real session.
-> Make a cozy isometric reading nook: a cut-away corner of a room with an armchair, a floor lamp glowing warm, a bookshelf full of colourful books, a round rug and a monstera, with evening sun through the window and a cat asleep on the rug.
+### Beyond code: Design
-**[Godogen](store/showcase/godogen/neon-drift.jpg)** · [Open harness](store/agents/godogen/)
+**[Blender](store/agents/blender/).** Ask for a lamp and the sliders that matter. Turn them and Blender rebuilds the geometry. Keep the versions you love.
-> Make a synthwave hoverbike racer: ride down a neon grid canyon toward a striped setting sun, weave between glowing pylons, hop barriers and collect energy cores, with a boost and three shields.
+<p align="center"><img src=".github/assets/readme/beyond/blender.gif" width="800" alt="Shape Lab in Blender: dragging height and twist sliders rebuilds a ribbon lamp, and chosen designs are kept."></p>
-**[Manim](store/showcase/manim/fourier-knight.jpg)** · [Open harness](store/agents/manim/)
+### Beyond code: Circuits
-> Draw a chess knight using nothing but spinning circles: a Fourier series of 120 epicycles, tip to tail, tracing its silhouette in gold.
+**[CircuitJS](store/agents/circuitjs/).** Build a filter. Change one resistor. Overlay the new trace, measure the difference and keep both.
-</details>
-<!-- store-showcase:end -->
+<p align="center"><img src=".github/assets/readme/beyond/circuitjs.gif" width="800" alt="Scope Lab in CircuitJS: an RC filter captured at 1 kΩ and 2 kΩ, traces overlaid and measured with cursors."></p>
-Every picture is real output from the harness's own toolchain, and the prompt is the one that made it.
+### Beyond code: Robotics
-## Coding agents, on every machine
+**[MuJoCo](store/agents/mujoco/).** Pin a moment in a robot's run. Shove it with 100 N. Watch two futures split.
-Keep the agents doing that work together, wherever they run.
+<p align="center"><img src=".github/assets/readme/beyond/mujoco.gif" width="800" alt="A Unitree Go2 in MuJoCo: the original and shoved futures play together with a height chart."></p>
-- **Real terminals that outlive the window.** Every agent runs in a persistent tmux session. Close the
- app and the agents keep working; if tmux goes down with a reboot, the daemon brings the panes back and
- resumes the sessions.
-- **Every engine, no wrappers.** Claude Code, Codex, Cursor, OpenCode, Pi, Hermes, Command Code, Devin,
- Muse Code, Amp, Kilo, Grok Build, Antigravity and GitHub Copilot. OpenHarness reads the transcript each
- agent already writes and installs the vendor's own hooks. Your credentials stay in `~/.claude`, `~/.codex` and
- so on. See the [engine list](docs/engines.md).
-- **All your machines in one window.** The laptop, the Mac mini at home, the server in the rack. Each
- runs a daemon with outbound connections only. Terminal traffic is end-to-end encrypted, the relay only
- forwards ciphertext, and it goes peer to peer over WebRTC when it can. No SSH server, VPN, or open
- port.
-- **Built for many agents at once.** Split panes, a keyboard-driven layout, fuzzy search across
- sessions and machines (**⌘O**), and one shortcut to the agents waiting on you (**⇧⌘I**). The
- [workspace guide](docs/app.md) and [keybindings](docs/keyboard.md) cover the rest.
+### Beyond code: Games
-<p align="center">
+**[Godogen](store/agents/godogen/).** Your agent makes a playable game. Miss a jump, rewind, try again. Pin the moment so the agent sees what you mean.
-<img width="1280" height="716" alt="ezgif-3355e1eae4366c76" src="https://github.com/user-attachments/assets/e3401aae-9402-4710-a46d-1da150ff96bf" />
+<p align="center"><img src=".github/assets/readme/beyond/godogen.gif" width="800" alt="Alpine Drift, a game made with Godogen: a run is rewound, retried and a moment is pinned with feedback."></p>
+### Beyond code: Music
-</p>
+**[Strudel](store/agents/strudel/).** The track is code you can perform. Bring voices in and out, mark the good parts, keep the WAV. [Hear it](https://github.com/user-attachments/assets/a3d4381b-5f55-406c-9d68-330cd8792fc5).
+<p align="center"><img src=".github/assets/readme/beyond/strudel.gif" width="800" alt="A live Strudel performance: voice lanes play beside the code, and the take is kept with markers."></p>
-### Run it
+### Beyond code: Chemistry
-1. [Download the desktop app](https://harness.autonomous.ai/desktop) for macOS or Linux.
-2. Sign in to a coding agent you already use, with your own subscription, API key, or local model.
-3. Press **⌘N**, pick an agent or a harness, a machine and a project, and start.
+**[RDKit](store/agents/rdkit/).** Turn a bond and watch the molecule move. Follow the real energy curve. Keep the pose worth a closer look.
-On another machine:
+<p align="center"><img src=".github/assets/readme/beyond/rdkit.gif" width="800" alt="A bond scan in RDKit: the molecule rotates through sampled poses along an MMFF94 energy curve."></p>
-```bash
-curl -fsSL https://harness.autonomous.ai/cli/install.sh | bash
-harness login
-harness start
-```
+### Beyond code: Documents
-Then **Machines → Link Machine** in the app.
+**[Typst](store/agents/typst/).** Your agent writes a real PDF. Circle a detail, quote a line, leave a note. The next draft answers it.
-macOS is the primary tested platform. Linux builds exist and feature parity is in progress; Windows is
-work in progress. Embedded live viewers require macOS; remote viewers also need a current Harness
-CLI on both machines. The app and daemon still need a Harness account to start;
-[account-free local use is tracked](docs/development.md#account-free-local-use).
+<p align="center"><img src=".github/assets/readme/beyond/typst.gif" width="800" alt="A Typst PDF under review: notes are pinned to an area and a sentence, then carried to the next draft."></p>
-<details>
-<summary><b>Build from source</b></summary>
+### Beyond code: Data
-For macOS, install Node.js 20+, tmux, Xcode, and Flutter 3.47+ / Dart 3.13+:
+**[Jev Sheets](store/agents/jev-sheets/).** Test a question on a few frozen rows before you ask the whole sheet. Compare two wordings side by side.
-```bash
-git clone https://github.com/autonomous-ai/openharness.git
-cd openharness
-(cd cli && npm ci)
-make install-cli
-cd desktop
-flutter config --enable-swift-package-manager
-flutter pub get
-flutter run -d macos
-```
+<p align="center"><img src=".github/assets/readme/beyond/jev-sheets.gif" width="800" alt="Question Lab in Jev Sheets: two wordings of a question are compared on frozen rows, recorded with practice data."></p>
-`make install-cli` installs this checkout's CLI and restarts the local daemon. The
-[development guide](docs/development.md) covers tests and isolated environments.
+## Domain-specific harnesses (DSH)
-</details>
+Code is the common medium. Geometry scripts make parts. Netlists make boards. Animation code makes film.
-### How it fits together
+A **domain-specific harness** turns a coding agent into a specialist. It brings the craft's
+instructions and skills, a pinned toolchain, a project template, checks and a **live viewer**.
+You chat on one side. The board, the part or the game takes shape on the other.
-```mermaid
-flowchart LR
- device["Harness device"] -- USB --> daemon
- app["Harness app<br/>(Flutter)"] -- loopback --> daemon["harness daemon<br/>(TypeScript)"]
- daemon --> tmux["tmux"] --> agents["Claude Code · Codex · OpenCode · …"]
- daemon --> dsh["DSH toolchain<br/>+ live viewer"]
- daemon <-- "E2EE · WebRTC" --> relay["Harness relay"]
- relay <--> remote["daemons on your<br/>other machines"]
-```
+The agent does the reasoning. The harness brings the tools and the view. It's a folder with a
+`harness.json`, so adding a craft never touches the app.
-The [architecture guide](docs/architecture.md) covers the daemon, the session model, transport and
-encryption.
+<!-- store-catalog:start -->
+### 49 harnesses in the Store
-## Your first DSH in ten minutes
+| Category | Agents and harnesses |
+|---|---|
+| **Coding** | [Claude Code, Codex, Cursor, OpenCode, Pi, Hermes, Command Code, Devin, Muse Code, Amp, Antigravity, GitHub Copilot, Grok Build, Kilo Code](docs/engines.md), [Harness Monitor](store/agents/harness-monitor/), [Machine Monitor](store/agents/machine-monitor/) |
+| Design | [Autonomous Workshop](store/agents/autonomous-workshop/), [Blender](store/agents/blender/), [Bonsai MCP](store/agents/bonsai-mcp/), [Creative Direction](store/agents/creative-direction/), [Excalidraw](store/agents/excalidraw/), [FreeCAD](store/agents/freecad/), [Generative Art](store/agents/generative-art/), [OpenSCAD](store/agents/openscad/), [text-to-cad](store/agents/text-to-cad/) |
+| Engineering | [Autonomous Circuit](store/agents/autonomous-circuit/), [CircuitJS](store/agents/circuitjs/), [Home Assistant](store/agents/home-assistant/), [KiCad](store/agents/kicad/), [Orca Slicer](store/agents/orca-slicer/), [Yosys](store/agents/yosys/) |
+| Media | [Comfy MCP](store/agents/comfy-mcp/), [Manim](store/agents/manim/), [OpenMontage](store/agents/openmontage/), [Remotion](store/agents/remotion/) |
+| Music | [Ableton AI](store/agents/ableton-ai/), [JUCE Agent Toolkit](store/agents/juce-agent-toolkit/), [Music Studio](store/agents/music-studio/), [Score](store/agents/score/), [Strudel](store/agents/strudel/) |
+| Productivity | [Jev Sheets](store/agents/jev-sheets/), [Marp](store/agents/marp/), [Typst](store/agents/typst/) |
+| Science & Data | [autoresearch-mlx](store/agents/autoresearch-mlx/), [Data Studio](store/agents/data-studio/), [Lab Bench](store/agents/lab-bench/), [marimo](store/agents/marimo/), [RDKit](store/agents/rdkit/) |
+| Simulation | [DimOS](store/agents/dimos/), [Drone Pilot](store/agents/drone-pilot/), [Foam-Agent](store/agents/foam-agent/), [MuJoCo](store/agents/mujoco/), [SimSkill](store/agents/simskill/) |
+| Games | [Game Master](store/agents/game-master/), [Godogen](store/agents/godogen/), [Phaser](store/agents/phaser/), [Voxel Worlds](store/agents/voxel-worlds/) |
+| Research | [Jev Browser](store/agents/jev-browser/), [Roundtable](store/agents/roundtable/) |
+| Local AI | [Grid](store/agents/autonomous-grid/), [MLX-LM](store/agents/mlx-lm/), [Ollama](store/agents/ollama/), [vLLM](store/agents/vllm/) |
-The [Hello World example](store/examples/hello-world/) is a Codex session that edits an HTML page
-shown in the shared Web Viewer:
+Upstream open-source tools and original workflows. 10 [shared viewers](store/viewers/) install alongside
+the harnesses that need them. Unlisted experiments are not shown.
+<!-- store-catalog:end -->
-```text
-hello-world/
- harness.json
- AGENTS.md # instructions for the agent
- template/index.html # copied into a new project
-```
+### Build your own
-Its manifest connects the pieces:
+The next harness is the one for your craft. Wrap a tool you love or your team's toolchain.
+It can live here or in your own repo. From this repo:
```json
{
@@ -252,45 +239,24 @@ Its manifest connects the pieces:
"id": "examples/hello-world",
"name": "Hello World",
"engine": "codex",
- "workspace": {
- "template": "template",
- "marker": "index.html"
- },
+ "workspace": { "template": "template", "marker": "index.html" },
"agent": { "instructions": "AGENTS.md" },
"viewer": { "use": "autonomous/web-viewer" }
}
```
-The CLI and the package protocol call a harness a DSH, so the commands are `harness dsh …`. From this
-checkout, with the `harness` CLI installed:
-
```bash
harness dsh install "$PWD/store/viewers/web-viewer" --link
-cp -R store/examples/hello-world ../my-first-harness
-harness dsh check ../my-first-harness
-harness dsh install ../my-first-harness --link
+cp -R store/examples/hello-world ../my-harness
+harness dsh check ../my-harness
+harness dsh install ../my-harness --link
```
-Press **⌘N → Hello World**, choose a new project, and ask it to “Say hello to Ada.” The agent edits
-`index.html`; the viewer reloads. Change `AGENTS.md` to try another workflow, then start a new session.
-`--link` keeps the package connected to your source directory; Store installs resolve viewer
-dependencies on their own.
-
-The [authoring guide](store/README.md) covers the full manifest, toolchains, live progress, verdicts,
-store pages, and publishing; the [package specification](store/spec/README.md) is the contract.
-
-| Piece | Responsibility |
-|---|---|
-| Engine | Runs the coding agent: Claude Code, Codex, OpenCode, and the rest |
-| Harness (DSH) | Packages a domain: engine, instructions, toolchain, workspace, and viewer |
-| Viewer | Shows and interacts with what the agent makes; shared across harnesses |
-| Session | One running agent, in one project, on one machine |
+Press **⌘N → Hello World** and say hello. The [authoring guide](store/README.md) covers the rest.
## Harness device
-<p align="center">
- <img src=".github/assets/hardware/answer.jpg" width="960" alt="A finger taps the round Harness device to answer an agent that redesigned the billing controls">
-</p>
+<p align="center"><img src=".github/assets/readme/device.gif" width="960" alt="A finger taps the round Harness device, speaks a task to fix the login flow, and the device shows the agent deploying, then a summary of the shipped fix."></p>
The optional **Harness device** is a round, always-on display that sits beside
your keyboard and shows your agents at a glance: what each one is doing, which one has finished, and
@@ -309,36 +275,10 @@ It is open hardware, all the way down. This repository has everything it takes t
[firmware guide](devices/harness-device/firmware/README.md) lists the supported boards and build
commands, and the [hardware guide](devices/harness-device/hardware/README.md) covers the design files.
-https://github.com/user-attachments/assets/97848065-61c6-40df-be66-a8247f69aa4c
-
-## Contributing
-
-There are three ways in, one for each layer of the stack.
-
-- **Make a DSH for a tool you use.** Start from Hello World, make one workflow work end to end, pin the
- toolchain, credit the upstream project, and show something worth looking at in the viewer.
-- **Improve the coding workspace.** Terminal behavior, engine support, the daemon, the relay, Linux and
- Windows.
-- **Hack the hardware.** Port the firmware to another board, remix the enclosure, add a feature to the
- device.
-
-Small fixes and notes about something that didn't work are welcome too. The
-[contribution guide](CONTRIBUTING.md) walks through each path.
-
-## Repository map
-
-```text
-desktop/ Flutter app and terminal workspace
-cli/ TypeScript CLI, daemon, engine adapters, and package runtime
-backend/ Relay and control plane
-store/ Domain-specific harnesses, shared viewers, registry, examples, and package spec
-provider/ Provider API contract, implementations, and conformance tests
-devices/ Harness device firmware, PCB, and enclosure
-```
+## Contribute
-- [Development and testing](docs/development.md) · [Extension points](docs/extending.md) · [CLI and automation](docs/cli.md)
-- [Desktop](desktop/README.md) · [Daemon](cli/README.md) · [Relay](backend/README.md) · [Provider API](provider/README.md)
-- [Security policy](SECURITY.md) · [License](LICENSE)
+Make a harness for a tool you love. Improve terminals, engines, the daemon or the relay. Port the
+firmware. Start with the [contribution guide](CONTRIBUTING.md).
-The repository is MIT licensed unless a folder says otherwise. Upstream tools, models, and assets keep
-their own licenses.
+[Development](docs/development.md) · [Extending](docs/extending.md) · [CLI](docs/cli.md) ·
+[Security](SECURITY.md) · [MIT license](LICENSE); upstream tools keep their own.
diff --git a/store/tools/presentation.mjs b/store/tools/presentation.mjs
index 9507357db..92e21463f 100644
--- a/store/tools/presentation.mjs
+++ b/store/tools/presentation.mjs
@@ -50,17 +50,12 @@ if (process.argv.includes('--categories-only')) process.exit(0);
const inventory = [
'<!-- store-catalog:start -->',
- '### Coding and beyond', '',
- `Start with a coding agent you already use. Explore ${agents.length} domain-specific harnesses when your`,
- 'next idea takes you further.', '',
+ `### ${agents.length} harnesses in the Store`, '',
'| Category | Agents and harnesses |', '|---|---|',
`| **Coding** | [${engines.map(([,name]) => name).join(', ')}](docs/engines.md)${codingHarnesses.length ? ', ' + codingHarnesses.map(harnessLink).join(', ') : ''} |`,
...craftGroups.map(c => `| ${c.name} | ${c.entries.map(harnessLink).join(', ')} |`), '',
- `These are the ${agents.length} harnesses currently listed in the Store catalog. They combine upstream`,
- 'open-source tools and original workflows, with instructions, setup, checks, and live views for each craft.', '',
- `The ${viewers.length} [shared viewers](store/viewers/) cover CAD, 3D models, documents, games, film, video,`,
- 'MuJoCo, web pages, isolated web previews, and studios. Viewer packages install alongside the',
- 'harnesses that need them. Experimental packages marked unlisted are not included above.',
+ `Upstream open-source tools and original workflows. ${viewers.length} [shared viewers](store/viewers/) install alongside`,
+ 'the harnesses that need them. Unlisted experiments are not shown.',
'<!-- store-catalog:end -->',
].join('\n');
let readme = read('README.md');
diff --git a/store/tools/showcase.mjs b/store/tools/showcase.mjs
index 6c7b033cb..11c6133d4 100644
--- a/store/tools/showcase.mjs
+++ b/store/tools/showcase.mjs
@@ -52,12 +52,6 @@ const gallery = [
].join('\n');
let readme = readFileSync(resolve(root,'README.md'),'utf8');
const marked = /<!-- store-showcase:start -->[\s\S]*?<!-- store-showcase:end -->/;
-if (marked.test(readme)) readme = readme.replace(marked,gallery);
-else {
- const start = readme.indexOf('<table>');
- const end = readme.indexOf('</table>',start);
- if(start<0||end<0) throw Error('Cannot find the README showcase table');
- readme = readme.slice(0,start)+gallery+readme.slice(end+'</table>'.length);
-}
-writeFileSync(resolve(root,'README.md'),readme);
+// The README no longer embeds the slideshow; update it only where the markers remain.
+if (marked.test(readme)) writeFileSync(resolve(root,'README.md'),readme.replace(marked,gallery));
console.log(`Saved ${examples.length} slides, ${examples.reduce((n,e)=>n+e.seconds,0)} seconds per loop, ${(statSync(join(output,'showcase.gif')).size/1048576).toFixed(2)} MiB. Preview frames: ${frames}`);