Your app, tested
end to end.
Your Claude or ChatGPT writes the tests. Assertify runs them on real browsers and catches what breaks.
Your Claude or ChatGPT writes the tests. Assertify runs them on real browsers and catches what breaks.
One platform instead of the QA stack you'd otherwise glue together
Assertify is an MCP server — the open protocol AI assistants use to drive other apps — so any agent that speaks it can drive Assertify: Claude, ChatGPT, your IDE, a local model in LM Studio or Ollama, or your terminal. Connect in ~30 seconds, no API keys. MCP is part of the paid cloud plan — talk to us; free on your own machine, you drive the same agents from Agent Chat inside the app.
Claude, ChatGPT, or Codex does the thinking from the flat plan you already pay for — Assertify only bills for run volume. Every other platform in this category has to price its own inference into your bill. That's the whole difference between $99 and $2,500.
$120k/yr Automation QA (SDET) · AI-QA platforms $1–2.5k/mo, enterprise overage past ~5k runs · Assertify $0 local, unlimited.
Everything a QA department does, working from the same context: what changed, what broke, and what to verify next.
Claude, ChatGPT, or Codex writes tests from your live app — plain English in, runnable cases out.
Upload your existing test cases or code tests — a CSV, a doc, or real code. Claude or ChatGPT imports each one into a runnable Assertify test.
Every test is an editable node graph. Branch flows, drop assertions, wire steps — no code required.
Click through your app once; Assertify captures the steps with real selectors and turns them into a clean test.
Download any test as real code — Playwright, Cypress, Selenium, WebdriverIO, Puppeteer or Robot Framework, in TypeScript, JavaScript, Python, Java or C#. No lock-in, ever.
When a locator breaks, the durable UI map re-resolves it against what the element actually is — deterministically, with no model in the loop. Available on local runs today.
Import Postman, OpenAPI, or a HAR. Assert on status, schema, and latency — then turn any collection into a load test.
Bugs are rejected unless they carry steps, expected, actual and evidence — then deduplicated by fingerprint, so the same defect is never filed twice.
No frameworks to learn, no selectors to write, no code changes to your app. Point it at a URL and it goes to work.
Open Agent Chat in the app — or add Assertify to Claude, ChatGPT, Cursor, or your IDE over MCP, live in ~30 seconds. Nothing to install in your app.
VS Code
Windsurf
Zed
LM Studio
Ollama
Hermes
Antigravity
Any MCP clientmcp.assertify.co
Point your agent at a URL. It explores the live product and writes real, grounded tests — happy paths, edge cases, validation. No selectors, no scripts.
Tests run on real browsers — real clicks, real APIs, never mocks — in our cloud, or free on your own machine with the local runner. Video and screenshots of every step, on demand or triggered from your pipeline.
Failures come back as evidence; your agent fixes the code and reruns. Healed and passing tests stay green — coverage compounds every release.
Claude, ChatGPT, or Codex does the authoring; alerts, tickets, and checks flow to the tools your team already lives in.
Assertify ships no AI of its own — Claude, from your plan, does the thinking in Agent Chat or over MCP.
Every test you author is also real, runnable code. Download it as a self-contained project for the framework and language your team ships — included with every plan, on top of the plain-English tests. The code is yours.
Two commands and Assertify runs tests in a browser on your own computer — authoring, exploration, self-healing, bug reports — for nothing. One more connects your own AI and unlocks Agent Chat: the whole loop, local.
Runs on your machine's own Chromium. Point your own Claude, ChatGPT, or a local Ollama model at it — nothing leaves your computer except the results you sync. Add assertify connect-model and the same runner powers Agent Chat in the app (see below).
Free to install, free to run. Create a free workspace and pair your machine in about a minute.
Chat is built into the app — your own Claude does the thinking, a real browser on your machine does the clicking. Your token never touches our cloud.
One paste — a Claude subscription token or an Anthropic / OpenAI API key, right in the app — and Chat lights up in the sidebar. assertify service install keeps the runner on in the background, reboots included.
browser_opendry_runcreate_testrun_testDone — “Coupon applies at checkout” is saved and green. It ran on this machine, so it cost nothing.
The desktop app bundles the local runner. Install, sign in, and your tests run on your own machine at $0 — everything syncs to your workspace.
Starting download…
On an Intel Mac? The app is Apple Silicon only — use the one-line install in Run local above, which fetches the Intel runner (macOS 10.15+).
Preview build — not notarized yet, so macOS shows “could not verify” the first time. Fastest fix: copy this, run it, then open the app.
Prefer no terminal? Open it once via System Settings → Privacy & Security → Open Anyway. Or skip Gatekeeper entirely with the one-line installer.
Preview build — not code-signed yet, so Windows SmartScreen shows “Windows protected your PC” the first time. Click More info → Run anyway to install.
Prefer no installer? Get it from PowerShell with the one-line installer.
Tests run on your own machine, so they cost us nothing and we charge you nothing — unlimited runs, every feature, no credit card. Want it off your machine? We set that up with you.
The whole product, on your machine
$0 /forever
Unlimited runs, no credit card
For when you want it off your machine
Let's talk
Set up with you — no self-serve checkout
Bigger team, SSO, or a security review? Get in touch.
Yes — it's live and self-serve at app.assertify.co. Sign up free (no credit card), install the desktop app, connect Claude Code, Codex, or Cursor in about 30 seconds, and start running tests right away. Free is unlimited and stays free, because the tests run on your own machine. Want the runs off your machine instead — cloud browsers, a shared team workspace, and CI? Talk to us.
Do — Assertify uses that same Claude or Codex to write the test. The difference is what happens on the fortieth run. Ask an agent twice and it can answer twice differently; an Assertify run is a replay with no AI in it, so a pass means the same thing every time. Its memory also lives on one laptop, so it re-finds the same bug every week and your teammates start from zero. Here the tests, runs and bugs are shared — and they keep working when you switch from Claude to ChatGPT.
No. Describe what to test in plain English and shape it on a visual canvas — branch flows, add assertions, wire the steps. Anyone on the team can author and maintain tests.
It drives a real browser to your URL — in our cloud, or free on your own machine with the local runner — and interacts with the live app the way a user would. For pages behind a login, add test credentials and it signs in for you.
Yes — one command (curl -fsSL https://assertify.co/install.sh | sh) installs a compiled local runner — no Python, nothing else to set up — that executes tests in a browser on your own machine at $0. assertify start pairs it with your workspace right in your browser; authoring, exploration, self-healing, and bug reports all run locally and sync to your workspace. One more command — assertify connect-model — connects your own AI and the same runner powers Agent Chat. It's web-only today (macOS, Linux, and Windows). See the Run local section above.
The AI writes the test once. After that there is no model in the run path at all — a run is a deterministic replay of recorded steps against real selectors the AI actually saw on your live app, so the same input gives the same verdict every time. That is the difference between a test and an opinion. When a test only passes on a retry, Assertify stamps it as such instead of showing you a clean green, so flakiness stays visible rather than being averaged away.
Assertify ships no AI of its own — it runs on the AI you already pay for, and you pick the model. Over MCP, your Claude, ChatGPT, or Codex plan connects directly: no API keys, no token bills. In Agent Chat, you connect a credential once with assertify connect-model — a Claude subscription token (via claude setup-token, no API key needed) or an Anthropic / OpenAI API key — and it stays on your machine. Either way we never resell you tokens: you pay your model provider, not us. Agent Chat is free; connecting your agent over MCP is part of the paid cloud plan.
No client needed — Chat is the first tab in the app, and it's free on every plan (MCP, which connects an outside client like Claude Code or Cursor, is part of the paid cloud plan). Install and pair the free local runner (see Run local above), start it (assertify start), connect your AI — paste a Claude token or API key right in the app, or run assertify connect-model — and ask for what you need: it authors and runs tests, files bugs with evidence, works heal requests, and /goal loops toward the coverage goal you set — all while streaming a live view of the browser it drives. Prefer it always on? assertify service install keeps the runner running in the background, reboots included. The reasoning runs on your own machine: your token and chat history go only to your model provider, never to the Assertify cloud, and the runs are free.
Yes — by running the Assertify runner inside your pipeline, on your own CI machine. There's a Linux build for exactly that, and Assertify exports a ready-made pipeline config for GitHub Actions, GitLab CI, Jenkins, CircleCI, and Azure Pipelines, plus the generated test code if you'd rather run it standalone. Runs happen on your hardware, not ours, and that's free. The paid cloud plan adds the other direction: a CI webhook that fires the run in our cloud, so nothing has to be installed on the build machine at all.
Ask your AI to verify a bug and Assertify reproduces it against your live app on a real browser, then records the verdict — with steps, a screenshot, and video — on the bug itself. Bugs are deduplicated by a deterministic fingerprint, so the same defect never gets filed twice, and a bug that comes back after being closed is filed as a new one so the regression stays visible.
No — we never use your tests, screenshots, or recordings to train models. Prompts only go to the AI model that runs the agent, and runs execute in isolated browsers. In Agent Chat the model calls happen from your own machine — your AI credential and chat history go only to your model provider, never to the Assertify cloud. See our Privacy Policy.
Your first AI-built test suite in a couple of minutes. Free to start, no credit card.