Repository navigation
Expand file tree
/
Copy pathcontent.ts
More file actions
663 lines (640 loc) · 26.1 KB
/
Copy pathcontent.ts
File metadata and controls
663 lines (640 loc) · 26.1 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
// The facts every page uses, in one place: the landing pages, the docs chrome, astro.config.mjs,
// robots.txt, the structured data and the social card (scripts/og.mjs) all read this file.
// Every command, package and claim comes from the repo, its docs or a public source linked
// here: no invented numbers, adopters, logos or testimonials.
//
// Copy rules (see the design guide): headlines run two beats, the second shown in the accent;
// ledes are one or two sentences; plain words, no "just" or "simply", no marketing adjectives.
import type { IconName } from '../components/site/icons';
export const site = {
name: 'agent-device',
/** The text in the accent wordmark block, the window tags and the giant closing wordmark. */
wordmark: 'agent-device',
/** The site origin, no trailing slash. Canonicals, the sitemap and robots.txt use it. */
url: 'https://agent-device.dev',
repo: 'https://github.com/callstack/agent-device',
/**
* What it is, in a few words. The home page title is "Name: subtitle" (src/site/meta.ts),
* 60 characters at most in all.
*/
subtitle: 'Open source device layer for agents',
/** The motto, in two beats. Used by the hero, the social card and the banner. */
headline: ['Open source device layer', 'for agents.'],
/** The hero lede: one or two sentences, naming who it is for. */
lede: 'Let coding agents control mobile, web, TV and desktop apps on devices and simulators, local or in the cloud.',
/** The meta description: 160 characters at most. */
description:
'Open source device layer for coding agents: control mobile, web, TV and desktop apps on devices and simulators, local or in the cloud, via CLI, MCP or Node.js.',
install: 'npm i -g agent-device',
/** Alt text for the social card (public/og-image.jpg): the lockup, the motto, the picture. */
ogAlt: 'agent-device by Callstack. Open source device layer for agents.',
/** Who holds the copyright in the footer: "© YEAR AGENT-DEVICE BY CALLSTACK". */
owner: 'Callstack',
callstack: 'https://callstack.com/?utm_source=agent-device&utm_medium=referral',
license: {
name: 'MIT',
url: 'https://github.com/callstack/agent-device/blob/main/LICENSE',
/** For the structured data. */
spdx: 'https://opensource.org/licenses/MIT',
},
npm: 'https://www.npmjs.com/package/agent-device',
discord: 'https://discord.gg/eYapw6F3',
/** Platforms or targets, in the order the brief fixes. */
platforms: [
'iOS',
'Android',
'HarmonyOS',
'tvOS',
'Android TV',
'Fire TV',
'Vega OS',
'macOS',
'Linux',
'Web',
],
/** The main language of the source, for the structured data. */
language: 'TypeScript',
/** No pixel mark: the lockup is the wordmark alone. */
mark: undefined as string[] | undefined,
};
/** Docs paths the pages link to. */
export const docs = {
intro: '/docs/introduction',
start: '/docs/quick-start',
install: '/docs/installation',
agents: '/docs/agent-setup',
snapshots: '/docs/snapshots',
replay: '/docs/replay-e2e',
debugging: '/docs/debugging-profiling',
clouds: '/docs/device-clouds',
proxy: '/docs/remote-proxy',
client: '/docs/client-api',
aiSdk: '/docs/ai-sdk',
eve: '/docs/eve',
sessions: '/docs/sessions',
capabilities: '/docs/capabilities',
security: '/docs/security-trust',
llms: '/llms.txt',
};
type Link = { text: string; href: string };
/**
* The nav links, mono and centred. The docs header shows the ones that leave the docs (any
* link not under /docs).
*/
export const nav: Link[] = [
{ text: 'Docs', href: docs.intro },
{ text: 'Showcase', href: '/showcase' },
{ text: 'Cloud', href: '/cloud' },
];
/** The footer row, right of the copyright, on every page and in the docs. */
export const footer: Link[] = [
{ text: 'Docs', href: docs.intro },
{ text: 'llms.txt', href: docs.llms },
{ text: 'Contact', href: '/contact' },
{ text: 'Discord', href: site.discord },
{ text: 'GitHub', href: site.repo },
{ text: `${site.license.name} license`, href: site.license.url },
{ text: 'Callstack', href: site.callstack },
];
/** Paths kept out of the sitemap. Pages rendered with `noindex` belong here too. */
export const noindexPaths = ['/404'];
/** The production trackers (components/site/Analytics.astro); both IDs are public in the page. */
export const analytics = {
googleAnalyticsId: 'G-0HE866FCRN',
commonRoomSiteId: 'b4585d93-5d45-4fff-bca4-177537b94e27',
};
/** The marquee under the hero: targets, then the app frameworks it drives. */
export const marquee = [...site.platforms, 'React Native', 'Expo', 'Flutter'];
/** The 404 line, in two beats. */
export const notFound = ['This screen never rendered.', `The rest of ${site.name} did.`];
/** Questions people ask, answered from the README FAQ. Inline `code` is allowed. */
export const faq = [
{
q: 'Is there an MCP server?',
a: 'Yes. `agent-device mcp` starts the official stdio MCP server. It is one entry point to the same runtime as the CLI and the typed Node.js API, so sessions, evidence and replay behave the same in all three.',
},
{
q: 'Does it work with React Native, Expo, Flutter and native apps?',
a: 'Yes. It drives native iOS and Android apps and React Native, Expo and Flutter apps on supported targets. The commands and evidence vary by target; `agent-device capabilities --platform <platform>` lists them.',
},
{
q: 'How is it different from Appium, Detox or Maestro?',
a: 'With agent-device an agent reads the app and chooses each command at run time. Appium, Detox and Maestro run suites people write and maintain. agent-device complements them: save a run as an `.ad` script or export it as strict Maestro YAML.',
},
{
q: 'Can it run in CI?',
a: 'Yes. Record a run as an `.ad` script, replay it in CI with `agent-device test`, and keep the screenshots and logs as artifacts.',
},
{
q: 'Can I build my own agent or QA product on it?',
a: 'Yes. The typed Node.js client is a public surface over the same runtime, with AI SDK tools and an Eve integration on top. TesterArmy e2e and T3 Code build on it.',
},
{
q: 'Can several agents share one machine?',
a: 'Yes. Sessions are scoped to the git worktree, and device claims stop parallel agents from taking over each other’s simulators and emulators. `agent-device device status` shows who holds what.',
},
];
/** One line of a terminal transcript: a command, its output, a comment, or a result. */
export type Line = { kind: 'cmd' | 'out' | 'add' | 'del' | 'dim' | 'ok'; text: string };
/**
* The hero: a prompt for each kind of job, the agent-device commands the agent runs for it, and
* the agent's reply with what it brings back. Prompts come from the README; commands and the
* output lines come from the README and the docs (the Contacts walkthrough, Debugging and
* profiling). Refs vary between runs.
*/
export const jobs = [
{
id: 'build',
name: 'Build',
prompt:
'Implement the onboarding screen, run it on the iOS simulator and Android emulator, and attach screenshots.',
transcript: [
{ kind: 'cmd', text: 'agent-device open MyApp --platform ios' },
{ kind: 'cmd', text: 'agent-device snapshot -i' },
{ kind: 'cmd', text: 'agent-device screenshot ./onboarding-ios.png' },
{ kind: 'cmd', text: 'agent-device close' },
{ kind: 'cmd', text: 'agent-device open MyApp --platform android' },
{ kind: 'cmd', text: 'agent-device snapshot -i' },
{ kind: 'cmd', text: 'agent-device screenshot ./onboarding-android.png' },
{ kind: 'cmd', text: 'agent-device close' },
] as Line[],
reply:
'The onboarding screen runs on both. Attached: onboarding-ios.png, onboarding-android.png.',
},
{
id: 'verify',
name: 'Verify',
prompt:
'Add a contact named Ada in the iOS Contacts app, verify it saved, and attach evidence.',
transcript: [
{
kind: 'cmd',
text: 'agent-device open Contacts --platform ios --save-script ./contacts.ad',
},
{ kind: 'cmd', text: 'agent-device snapshot -i' },
{ kind: 'out', text: '@e2 [button] "Add"' },
{ kind: 'cmd', text: 'agent-device record start ./recording.mp4' },
{ kind: 'cmd', text: 'agent-device press @e2 --settle' },
{ kind: 'add', text: '+ @e7 [text-field] "First name"' },
{ kind: 'cmd', text: 'agent-device fill @e7 "Ada" --settle' },
{ kind: 'add', text: '+ @e14 [text-field] "Ada"' },
{ kind: 'cmd', text: 'agent-device press \'role=button label="Done"\' --settle' },
{ kind: 'cmd', text: 'agent-device wait text "Ada"' },
{ kind: 'cmd', text: 'agent-device screenshot ./contact-saved.png' },
{ kind: 'cmd', text: 'agent-device record stop' },
{ kind: 'cmd', text: 'agent-device close' },
] as Line[],
reply:
'Ada is saved: her contact card shows after Done. Attached: contact-saved.png, recording.mp4, and contacts.ad to replay it.',
},
{
id: 'profile',
name: 'Profile',
prompt: 'Check whether this change causes unnecessary React Native re-renders.',
transcript: [
{ kind: 'cmd', text: 'agent-device open MyApp --platform ios' },
{ kind: 'cmd', text: 'agent-device react-devtools profile start' },
{ kind: 'cmd', text: 'agent-device fill \'id="catalog-search"\' "tart" --delay-ms 100' },
{ kind: 'cmd', text: 'agent-device react-devtools profile stop' },
{
kind: 'out',
text: 'React profile 8.3s: 8 commits, 1303ms rendering, 8 over 16ms (worst #6 188ms)',
},
{ kind: 'add', text: '1. memo-defeated ContactRow: memo bypassed on 480/480 updates' },
{ kind: 'cmd', text: 'agent-device react-devtools profile report ContactRow' },
] as Line[],
reply:
'Yes: ContactRow re-renders on every keystroke, its memo bypassed on 480 of 480 updates. The report names the props to stabilize.',
},
];
/**
* Who uses agent-device. Every entry is backed by a public source, checked on 2026-10-27:
* code in the adopter's own public repo, or a post they or Callstack published. Featured
* adopters show their official mark (`logo`) on the home page. Do not add entries without a
* source (the README lists some that have none).
*/
export type Adopter = {
name: string;
/** The project, when it isn't the company's main product. */
project?: string;
how: string;
sources: Link[];
/** Shown in the home page strip. */
featured?: boolean;
/**
* The official mark, in docs/public/logos. `mono` draws it as a white silhouette; `keep`
* keeps its own light and dark parts (a mark on a solid tile), in greyscale.
*/
logo?: { src: string; mode: 'mono' | 'keep' };
};
export const adopters: Adopter[] = [
{
name: 'Expensify',
logo: { src: '/logos/expensify.svg', mode: 'mono' },
how: 'An agent skill reproduces each bug, records before and after video, replays an `.ad` script, measures Sentry spans and profiles React renders, as evidence on the pull request.',
sources: [
{
text: 'agent-device-evidence skill',
href: 'https://github.com/Expensify/App/blob/main/.claude/skills/agent-device-evidence/SKILL.md',
},
{
text: 'Case study',
href: 'https://www.callstack.com/blog/how-expensify-uses-agent-device-for-mobile-bug-evidence-and-profiling',
},
],
featured: true,
},
{
name: 'Shopify',
logo: { src: '/logos/shopify.svg', mode: 'mono' },
project: 'FlashList',
how: 'An agent-device skill, wired into the repo’s agent instructions and its fix-a-GitHub-issue workflow.',
sources: [
{
text: 'Skill',
href: 'https://github.com/Shopify/flash-list/blob/main/.claude/skills/agent-device/SKILL.md',
},
],
featured: true,
},
{
name: 'MetaMask',
logo: { src: '/logos/metamask.svg', mode: 'mono' },
project: 'MetaMask Mobile',
how: 'A dependency of the mobile app, with a setup guide and an Android snapshot backend for its LLM workflows.',
sources: [
{
text: 'Setup guide',
href: 'https://github.com/MetaMask/metamask-mobile/blob/main/docs/readme/agent-device.md',
},
{
text: 'snapshot-backend.ts',
href: 'https://github.com/MetaMask/metamask-mobile/blob/main/tests/llm-workflow/android/snapshot-backend.ts',
},
],
featured: true,
},
{
name: 'Uniswap',
logo: { src: '/logos/uniswap.svg', mode: 'mono' },
project: 'Uniswap mobile',
how: 'Connects the mobile app to AWS Device Farm devices through agent-device.',
sources: [
{
text: 'deviceFarmConnect.ts',
href: 'https://github.com/Uniswap/interface/blob/main/apps/mobile/scripts/deviceFarmConnect.ts',
},
],
featured: true,
},
{
name: 'Expo',
logo: { src: '/logos/expo.svg', mode: 'mono' },
how: 'Expo Simulators, the remote iOS Simulators and Android Emulators on EAS, can be driven with agent-device, and Expo’s docs have an agent-device page. Agents also verify Orbit on physical iPhones, iPads and Android devices with it.',
sources: [
{
text: 'Expo Simulators docs',
href: 'https://docs.expo.dev/preview/simulators/introduction/',
},
{ text: 'agent-device and Expo', href: 'https://docs.expo.dev/agents/agent-device/' },
{ text: 'Orbit CLAUDE.md', href: 'https://github.com/expo/orbit/blob/main/CLAUDE.md' },
{
text: 'EAS Workflows QA agent on expo.dev',
href: 'https://expo.dev/blog/build-an-ai-qa-agent-for-expo-apps-with-eas-workflows-in-minutes-today',
},
],
featured: true,
},
{
name: 'T3 Code',
logo: { src: '/logos/t3-code.svg', mode: 'keep' },
how: 'Simulator and emulator support in T3 Code is built on agent-device.',
sources: [
{ text: 'Pull request #10677', href: 'https://github.com/pingdotgg/t3code/pull/10677' },
],
featured: true,
},
{
name: 'TesterArmy',
logo: { src: '/logos/testerarmy.png', mode: 'mono' },
project: 'e2e',
how: 'The mobile engine of the e2e testing agent, `@e2e-dev/mobile`, runs on agent-device.',
sources: [
{
text: 'packages/mobile',
href: 'https://github.com/tester-army/e2e/blob/main/packages/mobile/package.json',
},
{ text: 'e2e.tester.army', href: 'https://e2e.tester.army' },
],
featured: true,
},
{
name: 'React Navigation',
how: 'A repo script drives the example app with agent-device for contributors and their agents.',
sources: [
{
text: 'scripts/agent-device.ts',
href: 'https://github.com/react-navigation/react-navigation/blob/main/scripts/agent-device.ts',
},
],
},
{
name: 'Storybook',
project: 'Storybook for React Native',
how: 'End-to-end checks of the addons across Expo, Expo Router and Re.Pack examples.',
sources: [
{
text: 'e2e/agent-device',
href: 'https://github.com/storybookjs/react-native/tree/HEAD/e2e/agent-device',
},
],
},
{
name: 'Rocket.Chat',
project: 'Rocket.Chat React Native',
how: 'End-to-end tests run on e2e, whose mobile engine drives simulators and emulators through agent-device.',
sources: [
{
text: 'e2e/README.md',
href: 'https://github.com/RocketChat/Rocket.Chat.ReactNative/blob/develop/e2e/README.md',
},
],
},
{
name: 'OneKey',
how: 'A UI verification skill with agent-device rules for React Native, and regression scenarios.',
sources: [
{
text: '1k-ui-verify skill',
href: 'https://github.com/OneKeyHQ/app-monorepo/blob/x/.skillshare/skills/1k-ui-verify/SKILL.md',
},
],
},
{
name: 'Margelo',
project: 'react-native-quick-crypto',
how: 'The example app’s end-to-end runner drives the app with agent-device.',
sources: [
{
text: 'e2e runner',
href: 'https://github.com/margelo/react-native-quick-crypto/blob/main/example/test/e2e/runner.mjs',
},
],
},
{
name: 'App & Flow',
project: 'stim',
how: 'An agent driver package for stim built on agent-device.',
sources: [
{
text: 'agent-driver-agent-device',
href: 'https://github.com/appandflow/stim/blob/main/packages/agent-driver-agent-device/package.json',
},
],
},
{
name: 'Legend',
project: 'Legend List',
how: 'agent-device is the agent runtime for the Legend apps.',
sources: [
{
text: 'docs/agent-runtime.md',
href: 'https://github.com/LegendApp/legend-apps/blob/main/docs/agent-runtime.md',
},
],
},
{
name: 'Akveo',
project: 'UI Kitten',
how: 'agent-device is a dev dependency, with agent instructions for driving the showcase app.',
sources: [
{
text: 'AGENTS.md',
href: 'https://github.com/akveo/react-native-ui-kitten/blob/master/AGENTS.md',
},
],
},
{
name: 'Kortix',
project: 'Suna',
how: 'The repo’s end-to-end skill uses agent-device for mobile setup and debugging.',
sources: [
{
text: 'e2e skill setup',
href: 'https://github.com/kortix-ai/suna/blob/dev/.agents/skills/e2e/references/setup.md',
},
],
},
];
/** The /cloud page: the same commands from a laptop to CI to remote devices. */
export const cloud = {
headline: ['Same commands.', 'Any device, anywhere.'],
lede: 'Your agent keeps the same agent-device workflow when it leaves your laptop: replay in CI, run from Linux against a Mac that hosts the simulators, or drive hosted devices in a device cloud.',
paths: [
{
icon: 'panel',
label: 'Local',
title: 'On your machine',
body: 'Simulators, emulators and physical devices on your Mac or Linux box. Sessions follow your git worktree, so parallel agents don’t take over each other’s devices.',
cmd: 'agent-device open MyApp --platform ios',
href: docs.sessions,
},
{
icon: 'flag',
label: 'CI',
title: 'On every pull request',
body: 'Replay saved `.ad` scripts with `agent-device test`, keep screenshots and logs as artifacts, and report to JUnit.',
cmd: 'agent-device test ./workflows',
href: docs.replay,
},
{
icon: 'network',
label: 'Remote',
title: 'On a device host',
body: 'Run the agent on Linux or in a cloud workspace and execute on the Mac or Linux host that has the devices, through the remote proxy.',
cmd: 'agent-device proxy --port 4310',
href: docs.proxy,
},
{
icon: 'cube',
label: 'Device clouds',
title: 'On hosted devices',
body: 'The same commands drive BrowserStack, AWS Device Farm and Limrun devices built in, TestMu AI through its provider plugin, and Expo Simulators on EAS.',
cmd: 'agent-device connect limrun --platform ios',
href: docs.clouds,
},
] as { icon: IconName; label: string; title: string; body: string; cmd: string; href: string }[],
providers: [
{ name: 'BrowserStack', href: '/docs/browserstack' },
{ name: 'AWS Device Farm', href: '/docs/aws-device-farm' },
{ name: 'TestMu AI', href: '/docs/testmu' },
{ name: 'Limrun', href: '/docs/limrun' },
{ name: 'Revyl', href: '/docs/revyl' },
{ name: 'Expo Simulators', href: '/docs/device-clouds#expo-simulators' },
],
team: {
headline: ['Need devices for your team?', 'Talk to Callstack.'],
body: 'We help teams connect agent-device to their CI and pull request reviews, set up shared iOS and Android device hosts, and run agent-driven QA.',
},
};
/** The home page, section by section, in page order. index.astro only lays these out. */
export const home = {
/** The strip under the hero: who uses it, by name, linking to /showcase. */
usedBy: { label: 'Used by teams at', href: '/showcase' },
/** The bottleneck: why verification is the step that slows shipping with agents. */
bottleneck: {
label: 'Why agent-device',
headline: ['Agents write the code now.', 'Verification is what slows shipping.'],
body: [
'Coding agents turn out changes faster than anyone can check them on a phone. A mobile change only proves itself on a simulator or a device, and that is the one place agents could not go.',
'So the checking falls back on people: review rounds to ask whether it was run, QA passes on test builds, regressions found by users. agent-device moves that check to the agent, with evidence attached.',
],
},
/** The delivery loop, and the step agent-device owns. */
loop: {
label: 'The loop',
headline: ['Where agent-device fits', 'in shipping with agents.'],
lede: 'The same loop runs on every change. agent-device is the step where the agent checks its work on the real app, and leaves proof for the people who ship it.',
/** Grid order: the top row runs left to right, the bottom row back right to left. */
steps: [
{ n: 1, title: 'Task', body: 'You, or an issue, describe the change.' },
{ n: 2, title: 'Code', body: 'The agent writes and reviews the change.' },
{
n: 3,
title: 'Verify',
body: 'The agent runs the app, checks the change on the device, runs the e2e tests and records what it saw.',
product: true,
},
{
n: 6,
title: 'Feedback',
body: 'Crashes, bugs and slow screens come back as the next task, with a repro.',
product: true,
},
{
n: 5,
title: 'Review and ship',
body: 'People review proof instead of a description, and merge.',
},
{
n: 4,
title: 'Evidence',
body: 'Screenshots, video, logs and a replay script land on the pull request.',
product: true,
},
] as { n: number; title: string; body: string; product?: boolean }[],
},
/** Three stages of the loop, each with a picture and a proof point. */
inner: {
label: 'While the agent builds',
headline: ['It checks its own work.', 'Before you see it.'],
lede: 'Ask for a change and the agent runs it on the simulator, reads the screen, taps through the flow and fixes what it finds. You get a change that already ran, and fewer rounds of review.',
proof: 'Expensify',
/** The files the Verify job in the hero produces. */
evidence: [
{ kind: 'Screenshot', file: 'contact-saved.png', from: 'screenshot' },
{ kind: 'Video', file: 'recording.mp4', from: 'record stop' },
{ kind: 'Replay', file: 'contacts.ad', from: '--save-script' },
],
link: { text: 'Set up your agent', href: docs.agents },
},
outer: {
label: 'On every pull request',
headline: ['Verification that runs itself.', 'On every pull request.'],
lede: 'Save a run that worked as an `.ad` script and replay it in CI with no model in the loop, or run a QA agent against each build. Reviewers get screenshots, logs and a JUnit report, on simulators or hosted devices.',
proof: {
text: 'The EAS Workflows QA agent template',
href: 'https://github.com/callstackincubator/eas-agent-device',
},
/** An excerpt of examples/test-app/replays/checkout-form.ad. */
file: 'checkout-form.ad',
script: `context platform=ios timeout=60000
env APP_TARGET="Agent Device Tester"
open "\${APP_TARGET}" --relaunch
click "label=\\"Form\\""
fill id="field-name" "Ada Lovelace"
fill id="field-email" "[email protected]"
click id="submit-order"
wait "Order summary" 5000`,
ci: 'agent-device test ./workflows --reporter junit:./tmp/junit.xml',
link: { text: 'Replay and E2E testing', href: docs.replay },
},
feedback: {
label: 'From production back to the agent',
headline: ['When something breaks,', 'the agent brings a repro.'],
lede: 'Hand the agent the crash or the slow screen. It reproduces it on the device, captures the logs and network from that moment, and profiles the renders, so the fix starts from evidence.',
/** The diagnosis `react-devtools profile stop` prints, from Debugging and profiling. */
finding: {
summary: 'React profile 8.3s: 8 commits, 1303ms rendering, 8 over 16ms (worst #6 188ms)',
kind: 'memo-defeated',
component: 'ContactRow',
detail:
'memo bypassed on 480/480 updates; new highlight (object) 480x, onSelect (function) 480x from ContactList',
fix: 'wrap those props in useCallback/useMemo in ContactList',
},
link: { text: 'Debugging and profiling', href: docs.debugging },
},
/** The stories under "Who uses it", by adopter name. */
stories: ['Expo', 'Expensify', 'T3 Code', 'TesterArmy'],
/**
* Watch and learn: talks on closing the loop with agents. Titles and channels as YouTube lists
* them (checked through its oEmbed endpoint on 2026-10-27).
*/
videos: [
{
id: 'hKeFaFedu1s',
title: 'The Missing QA Loop for Mobile Agents',
who: 'Interview with Michał Pierzchała',
where: 'Callstack',
},
{
id: 'W2rHuUEo9rk',
title: 'Keynote',
who: 'Mike Grabowski and Michał Pierzchała',
where: 'Agent Conf 2026',
},
{
id: 'gxgmEOXcNBQ',
title: 'Automating Mobile QA with Cloud Agents',
who: 'Michał Pierzchała',
where: 'next app devCon 2026',
},
{
id: 'kZDU-k5r9kE',
title: 'Giving AI Agents Hands: Mobile Feedback Loops with Agent Device',
who: 'Mike Grabowski',
where: 'React Summit 2026',
},
{
id: 'r5P0detC4bs',
title: 'Cloud agents that test mobile apps on real devices',
who: 'Vercel',
where: 'Ship 26 London',
},
{
id: 'n7fO3NFaT8M',
title: 'Agent Device: AI agents meet mobile automation',
who: 'Callstack livestream',
where: 'Callstack',
},
],
/** Reading: articles and videos from the README. */
reading: [
{
text: 'How Expensify uses agent-device for bug evidence and profiling',
href: 'https://www.callstack.com/blog/how-expensify-uses-agent-device-for-mobile-bug-evidence-and-profiling',
},
{
text: 'Build an AI QA agent for Expo apps with EAS Workflows',
href: 'https://expo.dev/blog/build-an-ai-qa-agent-for-expo-apps-with-eas-workflows-in-minutes-today',
},
{
text: 'Building mobile QA agents with Vercel Eve',
href: 'https://www.callstack.com/blog/building-reviewable-mobile-qa-agents-with-vercel-eve',
},
{
text: 'How we optimized agent-device for mobile app automation',
href: 'https://www.callstack.com/blog/how-we-optimized-agent-device-for-mobile-app-automation',
},
],
};