The AI QA tester that proves every pass.

Describe an iOS test in one sentence. Every pass quotes what the screen showed.

Star on GitHub Join the cloud waitlist

Proof on every pass. Replays in seconds. MIT open source.

The test

  1. Search Wikipedia for Ada LovelaceSeen on screen · step 5“Ada Lovelace, English mathematician (1815–1852)”
    done
  2. Open the article about the English mathematicianSeen on screen · step 7“English mathematician (1815–1852)”
    done
  3. Verify it says she was born on 10 December 1815Seen on screen · step 7“10 December 1815 – 27 November 1852”
    done

PASS7 steps · 137 s

Recorded run, sped up

The Wikipedia app on Ada Lovelace's article. Tamam passed the test after quoting her birth date from the screen.

Runs on

  • iOS Simulator
  • Real iPhone
  • AWS Device Farm
  • BrowserStack
  • LambdaTest
  • Sauce Labs

Android and web: coming

How it works

From one sentence to a verdict you can check.

  1. “Search Wikipedia for Ada Lovelace, open the article about the English mathematician, and verify it says she was born on 10 December 1815”

    You write one sentence.

    Plain English, the way you'd brief a teammate. No code, no selectors.

    • Search Wikipedia for Ada Lovelace
    • Open the article about the English mathematician
    • Verify it says she was born on 10 December 1815

    It becomes a checklist.

    Each part of your sentence turns into a check that needs its own proof.

  3. Wikipedia search results for Ada Lovelace, with the agent's tap on the first result marked. Tap “Ada Lovelace”

    Step 5 of 7

    An agent uses your app.

    It taps, types and scrolls like a person, reading the screen as it goes.

  4. The article text, with “; 10 December 1815 – 27 November 1852” highlighted.
    “10 December 1815 – 27 November 1852”PASS · 7 steps · 137 s

    Every tick quotes the screen.

    A check passes only when it can cite the text your app showed.

Passes you can check, and reruns in seconds.

Every pass shows you what it saw. Passing tests get faster.

Every pass shows its proof.

Tamam ticks a check only when it can quote your app's screen. Here it added two items, opened the cart and read “Total: $35” before calling it a pass. No quote, no pass.

Tamam ticks a check only when it can quote the screen. No quote, no pass.

Step 3
Step 3: the Wireless Mouse page reads “Added to cart”.
Step 6
Step 6: the USB Cable page reads “Added to cart”.
Step 7
Step 7: the Cart reads “Total: $35”, highlighted as the evidence.
  1. Add the Wireless MouseSeen on screen · step 3 Added to cart
  2. Add the USB CableSeen on screen · step 6 Added to cart
  3. Open the CartSeen on screen · step 7 Cart
  4. Total is $35Seen on screen · step 7 Total: $35

PASS 7 steps · Shoply, cart test

Passing tests get faster and cheaper.

When a passing test compiles into a script and agrees with the AI, reruns replay it with no AI driving the steps. In our cloud (coming soon) that happens after one agreeing run; in the open-source version it's opt-in. When your app changes and the script no longer fits, the AI steps back in on its own.

Reruns replay as a script, with no AI driving the steps.

Same test: 23.1 s with the AI, 13.2 s as a replay. In our cloud (coming soon) it's on by default; in the open-source version it's opt-in.

Same test, two runs · shown about 17× faster than real time

Shoply test “Orders shows empty state”, timed start to finish by Tamam.

Same test, desk to device farm.

Write the test once. Run it on the iOS Simulator, your own iPhone, or real iPhones on AWS Device Farm, BrowserStack, LambdaTest and Sauce Labs.

Write the test once. Run it from your Mac, on device farms, or soon in our cloud.

Android and web next.

“Search Wikipedia for Ada Lovelace…”One test, unchanged

From your Mac

  • iOS Simulator
  • Real iPhone

Device farms

  • AWS Device Farm
  • BrowserStack
  • LambdaTest
  • Sauce Labs

Same sentence

  • Your Mac
  • Device farms
  • Our cloud soon

Every run is a ready bug report.

  • Screen recording
  • A screenshot of every step
  • The quoted evidence
  • Repro steps when it fails
  • Live view while it runs

Run it free today, or let us run it.

Same AI QA tester, same evidence on every pass. Pick where it runs.

Yours to run, today.

  • Free and MIT licensed. Read it, fork it, ship it.
  • Runs from your Mac, on the simulator or a real iPhone.
  • Bring your own Anthropic API key.
git clone https://github.com/tamamhq/tamam
Star on GitHubQuick start

Upload a build. We run the rest.

  • No setup, no machines to babysit. Upload your app and run your whole suite in parallel.
  • One dashboard for the team: watch runs live, replay the recording, share the evidence.
  • Fast replay on by default. Your suite gets faster and cheaper the more you run it.

Early access opens to a small group first. We'll email you when it's your turn.