Point DeckTest at your site or app and it tests the whole flow like a real person — signs up, reads the verification email, pulls the OTP, types it in, clicks through, and confirms the result three ways: on screen, in the DOM, and in the console. It's a real browser adapting live, not a brittle recorded script. Bring your own Claude — no extra AI fees.
DeckTest is a Mac app plus a small bridge extension for Chrome. Your Claude connects over MCP and drives the real browser on your machine. Setup is once, about five minutes — the quickstart guide walks every click.
Download the DMG, drag Deck Test to Applications, open it, and paste your license key. Signed for macOS.
One-time Chrome step: chrome://extensions → Load unpacked → the folder the app points you to. This is how Claude sees and clicks real pages.
One click adds DeckTest to Claude Desktop — or copy the config for Claude Code. Then ask: "test the signup flow at my-site.com."
A screenshot alone lies — a form can look fine while the request silently 500s. DeckTest confirms each step against what you see, what the page actually is, and what the browser is really doing underneath.
Looks at the rendered page the way a person would — is the success state there, did the modal close, is the button gone, does the dashboard actually load?
Reads the real elements, text and state — the right heading, the row that should exist, the field that should be cleared — not just pixels that happen to match.
Watches the browser console and every network call — a red error, a failed request, a 4xx/5xx the UI swallowed. The bugs a screenshot can't see.
Verification is where most test bots fall over. DeckTest doesn't. It fills the form, waits for the email to arrive, opens it, pulls the OTP or magic-link, and types the code in like a person would — then keeps going into whatever's behind the wall. No stubbed inbox, no skipped step.
When something fails, you don't get a red X and a shrug. DeckTest writes a QA report: the step that failed, a screenshot at the moment it broke, the console error or failed request behind it, and the exact steps to reproduce — so it's fixable, not just flagged.
Same BYO-key, local-first philosophy as the rest of the deck: no new account, no metering, no markup. The browser runs on your machine and the report stays on your machine.
Drives actual Chrome — real clicks, real typing, real navigation — and adapts when the page changes. No flaky selectors to re-record every time you ship.
Reads the verification email, pulls the code or magic-link, and types it in — the step that kills most test automation is the one DeckTest is built around.
Every run ends in a written QA report: failing steps named, screenshots at the moment of failure, console/network evidence, numbered repro steps.
Save a flow as a plain-text scenario and re-run it any time — same steps, fresh judgment. Great for smoke-testing before every deploy.
DeckTest is an MCP server, so it works with Claude Desktop AND Claude Code. One-click connect for Desktop, copy-paste config for the CLI.
The thinking happens on the Claude subscription you already have. DeckTest never sees, stores, meters or marks up your usage — no extra AI fees, ever.
No subscription, no seats to manage, no AI usage bill from us. Pay once, test forever.
Any website or web app you can reach in a browser — signup and login, checkout, onboarding, forms, dashboards. It goes through the flow end-to-end like a user: filling fields, reading verification emails, pulling OTPs or magic-links, clicking through, and confirming the result visually, in the DOM, and in the console + network.
Yes. DeckTest connects the Claude you already have — Claude Desktop or Claude Code — to a real browser over MCP. The thinking runs on your own Claude subscription; DeckTest never sees, stores, meters or marks up your usage. Beyond the $29 purchase there are no fees from us.
Both. For Claude Desktop, the app has a one-click "Connect to Claude Desktop" button that writes the MCP entry into your config. For Claude Code, there's a "Copy config" button — paste it into your MCP config or add it with one CLI command. The quickstart guide shows both.
No. It drives a real browser and adapts as it goes — reasoning about the page in front of it rather than replaying brittle selectors. When your UI changes, it adjusts instead of breaking. You can still save flows as YAML scenarios and re-run them — same steps, fresh judgment each run.
macOS 11 and newer, with Google Chrome. The current build is Intel; Apple Silicon Macs run it via Rosetta 2 (automatic, one click). A native arm64 build and a Windows version are planned — free updates within v1.x.
No trial — instead there's a 7-day, no-questions refund. Buy it, run it against your real app, and if it's not for you, email info@deckspace.dev from the address you paid with and a real person refunds you promptly.
Two Macs per license, tied to your key. Switching machines is self-serve — deactivate the old Mac in the app and activate the new one. Teams: one license per person.
The browser, the run, the screenshots and the QA report all stay on your Mac. Your Claude account talks to Anthropic directly under your own credentials. DeckTest has no account system and no telemetry. See privacy.
Download DeckTest for macOS, connect the Claude you already have, and let it test the flow like a real person would — then read the bug report.
DeckSpace