[{"data":1,"prerenderedAt":49},["ShallowReactive",2],{"article-en-qaptly-v0175-codex-support":3},{"article":4,"related":18,"prevArticle":38,"nextArticle":42,"availableLocales":46},{"id":5,"slug":5,"title":6,"excerpt":7,"content":8,"image":9,"date":10,"published":11,"status":12,"category":13,"keywords":14,"meta_description":15,"og_image":16,"custom_css":16,"created_at":10,"updated_at":17},"qaptly-v0175-codex-support","Qaptly Desktop now supports Codex","From v0.17.5, the AI Assistant in Qaptly Desktop officially supports Codex. If you already have a ChatGPT subscription, you can now have the AI Assistant use Codex and save on costs in day-to-day test design and automation work.","\u003Cp>The AI Assistant in Qaptly Desktop now finally supports CLIs other than Claude Code.\u003C/p>\u003Cp>Last year we released a version built on the Claude Agent SDK, so users could use the AI Assistant directly in Mobile Test Designer to organise test cases, check test targets, confirm device status, or quickly work out the next step while designing an automation flow.\u003C/p>\u003Cp>Our team has always used GitHub Copilot, Claude Code and Codex, so our internal builds support all of them.\u003C/p>\u003Cp>This time we decided to release Codex support officially.\u003C/p>\u003Cp>The main reason is that users have recently been asking about usage limits on the Agent SDK under a Claude subscription.\u003C/p>\u003Cp>For us this is straightforward. Qaptly's AI Assistant exists to save SQA time. If usage limits on the AI Assistant start making users worry about cost or quota, then supporting only Claude becomes a problem in itself.\u003C/p>\u003Cp>If you already have a ChatGPT subscription and Codex installed on your machine, you can now have the Qaptly Desktop AI Assistant use Codex. That should save you some usage and cost in day-to-day test design and automation work.\u003C/p>\u003Ch2>How to set it up\u003C/h2>\u003Cp>Setup is simple:\u003C/p>\u003Cp>Go to Settings → AI to find the AI Assistant settings.\u003C/p>\u003Cp>The current options are:\u003C/p>\u003Cul>\u003Cli>\u003Cp>Auto\u003C/p>\u003C/li>\u003Cli>\u003Cp>Claude Code\u003C/p>\u003C/li>\u003Cli>\u003Cp>Codex\u003C/p>\u003C/li>\u003C/ul>\u003Cimg class=\"rounded-lg shadow-lg mx-auto block max-w-full h-auto\" src=\"https://cdn.sqa.tw/articles/1778937209593-k6rs0pea1gq.jpg\" alt=\"qaptly_desktop_codex_setting.jpg\">\u003Cp>One thing to note: if both Claude Code and Codex are installed on your machine, select Codex manually.\u003C/p>\u003Cp>Auto currently prefers Claude by default. If you want to save on Claude Agent SDK usage, or would rather use your ChatGPT / Codex quota, don't choose Auto; select Codex directly.\u003C/p>\u003Ch2>What the AI Assistant can do in Mobile Test Designer\u003C/h2>\u003Cp>Once enabled, a floating AI Assistant appears in Mobile Test Designer. You can ask questions right from the current test design screen, for example whether a device is available, whether the test target is complete, or ask the AI to help break down test cases and automation steps.\u003C/p>\u003Cp>The point is not to hand all testing work over to AI.\u003C/p>\u003Cp>A closer description is this: the AI completes the work of automating test cases for you while you watch. It operates the app and verifies results through Qaptly, and builds the test case step by step.\u003C/p>\u003Cp>Qaptly has never been about packaging AI as a magic feature. The aim is to bring it into the actual SQA workflow and cut down the repetitive, fragmented work that still cannot be ignored. This matches how we have always positioned Qaptly: let the tool handle tedious test execution and process housekeeping, so SQA can spend their time on the quality judgements that matter more.\u003C/p>\u003Cimg class=\"rounded-lg shadow-lg mx-auto block max-w-full h-auto\" src=\"https://cdn.sqa.tw/articles/1778937189628-1i60bq1h774.jpg\" alt=\"qaptly_desktop_support_codex.jpg\">\u003Ch2>GitHub Copilot CLI is on the way\u003C/h2>\u003Cp>If you have auto-update turned on, you may receive a build with GitHub Copilot CLI support. Once it passes our external test scenarios, we will open the feature to everyone in the next release.\u003C/p>\u003Cp>If you have already updated to Qaptly Desktop v0.17.5, go to Settings → AI and turn on the AI Assistant. If you see a GitHub Copilot option, your version already has it.\u003C/p>","https://cdn.sqa.tw/blog_pictures/qaptly_support_codex.jpg","2026-05-16T13:01:00+00:00",true,"published","sqa","Qaptly Desktop, Codex, AI Assistant, Mobile Test Designer, ChatGPT, Claude Code, test automation, SQA","Qaptly Desktop v0.17.5 adds Codex support. Besides Claude Code, you can now pick Codex in Settings → AI so the AI Assistant helps with test design and automation.",null,"2026-05-16T14:04:13.318+00:00",[19,26,32],{"id":20,"slug":20,"title":21,"excerpt":22,"image":23,"date":24,"category":25},"ai-development-sdlc-release-verification","When AI builds it in two hours but verification still takes three days","Once AI speeds up development, spec sign-off, testing and release do not automatically speed up with it. Using a real case from a booking system SaaS, we describe how we adjusted the SDLC to cut full release verification from two or three days to under two hours, and how the SQA role changed along the way.","https://cdn.sqa.tw/articles/ai-sdlc-verification-bottleneck-cover.jpg","2026-11-05T12:00:00+08:00","perspective",{"id":27,"slug":27,"title":28,"excerpt":29,"image":30,"date":31,"category":25},"sme-guide-to-choosing-ai-agent","AI agents are everywhere. How should a small business choose one?","We were recently asked to run a full post-deployment review of an AI agent, and found it had quietly caused five times too much raw material to be ordered from customer order emails, and nearly created a liquidity risk on the finance side. The problem with agents is rarely that they are not smart enough. It is that nobody asked the right questions.","https://cdn.sqa.tw/articles/sme-agent-black-box-inspection.jpg","2026-09-01T15:00:00+00:00",{"id":33,"slug":33,"title":34,"excerpt":35,"image":36,"date":37,"category":25},"ai-coding-team-collaboration-three-patterns","Three patterns we see once AI coding reaches the team","After working with more than ten development teams over the past six months, we keep seeing the same thing: individuals getting faster with AI does not mean the team is ready to get faster together.","/images/blog/ai-coding-teams-editorial.jpg","2026-08-20T08:00:00+00:00",{"slug":39,"title":40,"date":41},"ai-agent-workflow-critical-things","Once AI joins the workflow, the things that matter most are the easiest to underestimate","2026-03-24T16:10:00+00:00",{"slug":43,"title":44,"date":45},"ai-needs-customer-insight-not-just-code","Two things we learned from a North American e-commerce project","2026-08-17T02:30:00+00:00",[47,48],"en","zh",1794970112860]