Your AI tests your app. Assay keeps the proof.
Connect Claude Code, Codex, Cursor or Hermes. It maps your app, writes the test cases, runs them on your simulators, emulators and browsers, and files bugs with a screenshot of every step. Your whole team sees the results in one place.
Release day still ends with someone tapping through the app.
Two people, forty screens, every release. Whatever sits at the end of the list gets skipped when it's late.
Hand-written scripts break when a button moves, and someone spends their week keeping them alive.
It can test the app, once. The steps, results and screenshots go away with the conversation, and nobody else on the team sees them.
Assay gives your AI a place to keep its work: a map of the app, test cases with every version, each run step by step, and the bugs it found. The next run starts from what the last one learned, and your team can check all of it.
Your AI does the testing. Assay remembers it.
Four steps, all from the AI you already use. Nothing to install on your devices: it drives the simulators, emulators and browsers you already have.
Join the pilotAdd one MCP server, then sign in from your browser. Each person connects their own AI; owners see every connection.
Your AI walks the app, saves each screen to the app map, then writes cases you review. A case names the test data it needs; the values stay on your machine.
Pick the phones, tablets, browsers and screen sizes that matter. Your AI runs each case on each of them and reports every step with a screenshot, so any run can be replayed tap by tap.
A failed step becomes a bug with the steps, what was expected, what the screen showed and the screenshots. When the fix ships, ask your AI to run the case again; a pass closes the bug with the run that proves it.
One place to see what your AI tested, for everyone who wasn't in the chat.
Developers, QA, product and the team lead see the same suite, the same runs and the same bugs.
Your AI saves each screen it finds and the control that leads there, so the next run starts from what it already knows.
Cases with every version and who changed them, and the last result on each phone, tablet and browser.
Each step with its screenshot, where the tap landed and what the AI expected to see.
Steps, expected and actual, screenshots, and a share link for people outside Assay.
The AI does the tapping. The record stays.
| Question | Assay | People tapping | Scripted UI tests | An AI in a chat |
|---|---|---|---|---|
| Who writes the tests | Your AI, and you review them | A QA team | Engineers, by hand | The AI, from scratch each time |
| Where results live | In Assay: every run, step and screenshot | A checklist | CI logs | The chat, until it scrolls away |
| Same steps each time | Yes: the AI follows the saved case | Depends on the day | Yes | No |
| When the UI changes | Your AI updates the case; old versions are kept | People adapt | The script fails until someone fixes it | It adapts, differently each time |
| Who can see it | The whole team, and a share link per bug | Whoever did it | Whoever reads CI | The person in the chat |
| AI cost | Your own AI plan | None | None | Your own AI plan |
Your AI tests the app. It never holds your secrets.
Cases name the data they need, like login_email or saved_address. The values live on your machine and never reach Assay.
test_data: login_email, saved_addressMark Delete account or Place order as Never tap. Assay refuses that step and stops the run, even when the AI asks.
Never tap · Delete accountNever tap · Place orderEach person signs their own AI in. Owners see every connection in the workspace and can block any of them at once.
Claude Code ✓ · Cursor ✓ · Hermes ✓ · Codex blockedThe audit log shows who did what and through which AI app, including calls Assay refused. Run screenshots are deleted after 30 days; results stay.
audit log · kept 1 yearFree while we run the pilot.
We're opening Assay to a small group of teams. Use it on every app you ship; nothing is metered during the pilot.
- 1Sign up with your work email.
- 2Enter your pilot access code to create your team's workspace.
- 3Connect your AI and ask it to map your app.
- Apps, cases and runs without limits
- iOS, Android, Flutter, React Native and web
- The whole team, with owners and members
- AI connection controls and the audit log
What teams ask before they connect.
Which AI do I need?
Any MCP client: Claude Code, Codex, Cursor, Hermes Agent, Claude or ChatGPT. Testing runs on your own AI plan, and Assay never sees your conversations.
Can I use a local model?
Yes, with Hermes Agent. The model runs on your machine, so prompts and code never reach an AI provider. Assay still keeps the cases, each step with its screenshot, and the bugs. Pick a model that can see images, like Gemma 4 31B or Qwen 3.5 27B, on a machine with 32 GB of memory or more. Runs take longer than on a cloud model.
Does Assay run my app?
No. Your AI drives the simulators, emulators and browsers on your machine or your CI runner. Assay keeps the app map, the cases, each step with its screenshot, and the bugs.
Which platforms does it test?
Native iOS and Android, Flutter, React Native and web apps. Web cases run in each browser and screen size you pick; a Flutter or React Native case runs on both phones.
Does Assay see our test data?
No. Cases name the data they need, like login_email. The values stay on your machine and the AI fills them in there.
What happens when the UI changes?
Your AI updates the case, and each edit is kept as a version you can compare or restore. A run that can no longer follow its steps is marked broken, not failed, so nobody chases a bug that isn't there.
Can it tap something it shouldn't?
Mark any control as Never tap, like Delete account or Place order. Assay refuses that step and stops the run, even when the AI asks.
What does it cost?
Nothing during the pilot, with no limits on apps, cases or runs. The AI's work counts against your own AI plan.
Can bugs go to our tracker?
Azure DevOps and GitHub are next. Until then every bug has a share link anyone can open without signing in.
Test your next build, and keep the proof.
Connect your AI in a few minutes, then ask it to map your app. Your team sees every case, run and bug as it lands.
Join the pilot Sign in