# Website QA Agent Template: Mobile and Desktop Checks | FlyMy.AI > Clone a website QA agent: it checks every link, opens each page in a real cloud browser on mobile and desktop, reads the screenshots for visible defects and publishes a report page with what to fix. Page: https://flymy.ai/templates/website-qa/ Agent template v1.2 Updated 2026-10-27 # Website QA - desktop and mobile Checks every link, opens each page in a real cloud browser on mobile and desktop, reads the screenshots for visible defects and publishes a report of what to fix. - Every link checked, broken ones listed by page - Each page in a real browser on mobile and desktop, its screenshots read for defects - A private report page with what to fix first [Clone this project](https://app.flymy.ai/projects/fleet%3A8ccc3fc9-72a8-4a79-ad67-afa514116cd0) [Use from Claude / Codex](https://flymy.ai/templates/website-qa/#start) By FlyMy.AI · templates/website-qa v1.2 Website QA - desktop and mobile and its tools: Cloud browser (Browser Use) (per session: provider cost plus the service margin; $0.009-0.083 per screen, measured on 3-6 pages with screenshots), Gemini vision (per token; about $0.006 per page with its mobile and desktop screenshots, measured 2026-10-27), Code execution ($0.0666 per sandbox hour, billed per second; about $0.0003 per run, measured), Report page (free; one private page on app.flymy.ai per run, $0 measured), Chat files (free; report.md, results.json and report.html in the chat, $0 measured) ## Proof - **12 / 12** browser checks, 6 pages on mobile and desktop, 36 screenshots read 2026-10-27 - **$0.21** a run on 6 pages and 2 screens with the visual check 2026-10-27 - **~3-5 min** from start to the report page 2026-10-27 - **Frozen + SDK** frozen version and Python SDK runs passed 2026-10-26 - **Clone run** a copy of the reference ran on its own 2026-10-26 - **[Public](https://app.flymy.ai/projects/fleet%3A8ccc3fc9-72a8-4a79-ad67-afa514116cd0)** reference project to clone **1** agent one QA agent **5** tools priced below **on demand** runs schedule after the first report **$0.21** per 6-page run measured ## What it does - Every link on the tested pages is checked over HTTP; robot blocks are listed apart, not as broken. - Each page opens in a real cloud browser at your screen size, scrolled to the end and measured. - Up to 3 screenshots of every page are read by Gemini 3.8 Flash for visible defects, linked in the report. - Read-only: no logins, no forms, no purchases. - The same script every run, checked by its SHA256, so weekly reports compare. **Ideal for** Site owners, marketers and developers who ship pages often and want a mobile check before or after each release. ### Connects to - **Cloud browser (Browser Use)** ~$0.01-0.08 / screen built in - **Gemini vision** ~$0.006 / page built in - **Code execution** ~$0.0003 / run built in - **Report page** free built in - **Chat files** free built in ## Project diagram from the reference project **The diagram as text** - **Inputs:** Website address, Pages: a number or your list, Screens: mobile to desktop - **Lead:** Website QA - desktop and mobile (gpt-6-luna) uses Cloud browser (Browser Use), Gemini vision, Code execution, Report page, Chat files - **Tool:** Cloud browser (Browser Use) - per session: provider cost plus the service margin; $0.009-0.083 per screen, measured on 3-6 pages with screenshots (built into FlyMy.AI) - **Tool:** Gemini vision - per token; about $0.006 per page with its mobile and desktop screenshots, measured 2026-10-27 (built into FlyMy.AI) - **Tool:** Code execution - $0.0666 per sandbox hour, billed per second; about $0.0003 per run, measured (built into FlyMy.AI) - **Tool:** Report page - free; one private page on app.flymy.ai per run, $0 measured (built into FlyMy.AI) - **Tool:** Chat files - free; report.md, results.json and report.html in the chat, $0 measured (built into FlyMy.AI) - **Output:** Private report page (from Report page) - **Output:** Report files (md, json, html) (from Chat files) ## Day one from the reference project - 01 **Clone** The QA agent with its tested script; nothing to connect. free - 02 **Answer once** Website, pages, screens, report and a weekly check, in one message. free - 03 **First report** Links over HTTP, one cloud browser per screen, the screenshots read, the report page. $0.21 for 6 pages x 2 screens - 04 **Weekly check** Optional: Schedule Agent under the report repeats it. the same per run ## How it runs templates/website-qa v1.2 Runs when you ask, in about 3-5 minutes. Choose Schedule Agent under a report to repeat it, for example every Monday at 09:00. ## The agent team from the reference project ### `Website QA - desktop and mobile` QA lead Finds the pages, checks every link, runs one cloud browser per screen, reads the screenshots and publishes the report, with one tested script. Cloud browser (Browser Use), Gemini vision, Code execution, Report page, Chat files **Instruction excerpt** Test the website on the chosen screens, at most the chosen number of pages, and give the owner a report page. Every check is read-only: no logins, no forms, no purchases. The work runs a tested script, loaded by its link and checked by its SHA256. Budget: your project budget on demand or weekly, gpt-6-luna, last run $0.214 on 2026-10-27 (6 pages of flymy.ai on mobile and desktop, 12 of 12 browser checks, 36 screenshots read) ## You will be asked once, in one message - **Which website?** The address to test; pages behind a login are skipped. `https://flymy.ai` - **How many pages, or which?** 1-15 from the sitemap or the homepage (default 6), your own list, or pages to always include on top. `6, always /pricing/` - **Which screens?** small-mobile 360, mobile 390, tablet 768 or desktop 1440 px. `mobile, desktop` - **Every issue in the chat?** The report page and three files come every time; the chat shows the top issues. `top issues` - **A weekly check?** Day, time and time zone, set with Schedule Agent after the first report. `no` Keys and passwords never go into the chat: you connect accounts on their own pages. ## What it costs Measured on 2026-10-27 **0%** template markup or purchase fee [Usage pricing →](https://flymy.ai/pricing/) **Price of one run** | Item | Unit price | Per run | Cost | |---|---|---|---| | Cloud browser (Browser Use), 2 screens | provider cost plus the service margin | 2 sessions, 6 pages each, up to 3 screenshots a page | ~$0.166 | | Gemini vision: the screenshots | per token | 1 read per page, mobile and desktop together | ~$0.036 | | Agent: gpt-6-luna and the run check | per token | about 9 model calls | ~$0.008 | | Code execution: links and report | $0.0666 per sandbox hour | 2 calls | ~$0.001 | | Report page and three chat files | free | 1 page, 3 files | $0 | | One run, 6 pages on 2 screens | $0.21 | | | Measured 2026-10-27 on the reference with the visual check: 6 pages of flymy.ai on mobile and desktop cost $0.214; 3 pages of docs.flymy.ai on 2 screens cost $0.047. Long pages take more browser time. Every run shows its itemised price. ## What stays yours the same for every template ### Your connections stay yours A copy never carries the author's keys or connections. ### Nothing runs until you answer Clone creates the agents and a setup chat; nothing bills before your reply. ### A budget you set 5 / 20 / 50 USD per day / week / month by default; a run cannot outspend it. ## Versions a copy keeps the version it was made from - v1.2 2026-10-27 Visual check: the cloud browser saves up to 3 screens of every page, Gemini 3.8 Flash reads them for visible defects, and the report shows each finding next to its screenshots. - v1.1 2026-10-27 Pages to always include on top of the picked ones; the result opens as fields with the report page and file links; a change asked later in the chat runs the tested script again. - v1.0 2026-10-26 First public template: link check over HTTP, one cloud browser per screen, a report page and three files, checksum-verified script hosted by FlyMy.AI. ## How to start 1 agent, 5 tools, your budget 01 - Clone in FlyMy.AI Open the reference project and press Clone; a setup chat asks once. [Clone this project](https://app.flymy.ai/projects/fleet%3A8ccc3fc9-72a8-4a79-ad67-afa514116cd0) 02 - Your Claude or Codex Connect FlyMy.AI with [skill.md](https://flymy.ai/skill.md) and paste the request below. 03 - API and SDK Agents MCP `projects_copy` with `fleet:8ccc3fc9-72a8-4a79-ad67-afa514116cd0`, or `client.projects.copy`. $ SETUP REQUEST ``` Read https://flymy.ai/templates/website-qa/template.txt and help me set up Website QA - desktop and mobile using my own connections. Ask for the missing inputs and budget before running it. ``` [Download brief](https://flymy.ai/templates/website-qa/template.txt) ## FAQ **Is this a working project or just an idea?** A working, public reference project. Clone it and its first chat asks five questions in one message; the first report page arrives in about 3-5 minutes. **Does it log in, fill in forms or buy anything?** No. It only reads pages: no logins, no forms, no purchases. Pages behind a login are not tested. **Which screens does it test?** small-mobile 360 px, mobile 390 px, tablet 768 px and desktop 1440 px; mobile and desktop by default. Each screen is its own real cloud browser. **What does it find?** Broken links (404, 410, 5xx or no answer), pages wider than the screen with the elements that stick out, broken images, tap targets under 24 px, a missing viewport tag and, from the screenshots, visible defects: overlapping or cut-off text, broken images or icons, empty areas, unreadable text, words that run together and a broken layout. Links that block robots (401, 403, 429) are listed apart. **What does the visual check look at?** The cloud browser saves up to 3 screens of every page on each screen size, and Gemini 3.8 Flash reads them for defects a visitor would notice, comparing mobile with desktop. It skips sticky headers over scrolled content, anything cut only by the edge of a screenshot, decorative elements and design taste. Each finding names the screen and the place, and the report links the screenshots so you can check it yourself. **What does a run cost?** Measured on 2026-10-27 with the visual check: $0.21 for 6 pages of flymy.ai on mobile and desktop (cloud browser $0.17, Gemini vision $0.04, the agent $0.01) and $0.05 for 3 docs pages. Long pages take more browser time. Every run shows its itemised price. **Can it check my site every week?** Yes. After a good report choose Schedule Agent under it and pick the day, time and time zone; every scheduled run publishes a new report page. **Illustrative example, not a live result** ``` site: https://example.com checked_at: screens: mobile 390 px, desktop 1440 px pages: 6, links checked: broken link: (404) on wider than the screen: on mobile ( px on px) - visual (mobile, screen 2): () screenshots read: report_page: ``` ## Sources [Reference project →](https://app.flymy.ai/projects/fleet%3A8ccc3fc9-72a8-4a79-ad67-afa514116cd0), [Browser Use in FlyMy.AI →](https://docs.flymy.ai/mcp/browser-use), [Code execution →](https://docs.flymy.ai/mcp/sandbox), [Usage pricing →](https://flymy.ai/pricing/) ## More templates Agent template ### [Instagram Trend Watcher](https://flymy.ai/templates/instagram-trend-watcher/) Finds the standout posts of the Instagram creators you pick and sends a daily digest with reuse ideas. Agent template ### [SEO / GEO Research Agent](https://flymy.ai/templates/seo-geo/) Crawls your site, checks your queries in Google and its AI Overview and publishes a report page with up to 10 actions, each with its evidence. Agent template ### [PromptLinks](https://flymy.ai/templates/promptlinks/) Two research agents find where your brand can earn links - a new PR article or an existing article - and publish one page with both tables. Content and connection guidance checked 2026-10-27. See linked documentation for current availability and prices. --- For agents: the FlyMyAI operating guide (sign up, connect, run models, tools and agents) is https://flymy.ai/skill.md, and https://flymy.ai/llms.txt lists the other pages and the documentation.