QA.tech
QA.tech is an AI agent testing platform for web and native mobile apps. Its agents run end-to-end, regression, exploratory and pull-request checks described in plain English, with no scripts, no selectors and no access to your source code.
What is QA.tech?
QA.tech is a software testing platform in which AI agents, rather than scripted suites, exercise an application the way a customer would. The vendor's formulation is test by intent, not brittle scripts: an agent is given a goal and works towards it, so a change in the interface does not automatically break the test. Because the agents read the screen visually, there are no selectors and no scripts to maintain, and the platform never needs access to the source repository, since it tests the product from the outside. Onboarding starts with a crawl: the agent explores the application and builds a feature graph of what it found, so no upfront configuration of pages or flows is required. From there tests are written in natural language in a chat, and each briefing adds to knowledge the agents keep from one run to the next. Coverage spans web, mobile web, iOS and Android apps. Announced test types include end-to-end UI, pull-request testing, dynamic and native mobile testing, cross-device, regression, API, email and SMS checks, and the site publishes 36 use cases sorted by type (7 end-to-end, 7 regression, 3 exploratory, 2 accessibility, 1 visual, 1 smoke) and by vertical (18 SaaS, 6 e-commerce, plus fintech, healthcare, marketplace and CRM). Less common capabilities are called out too: opening an email, reading an SMS code, nested iFrames, two agents acting as two users in isolated sessions, and tunnels into a local environment. The platform sits inside the delivery workflow. A GitHub or GitLab bot reads pull-request context, such as notes, comments and Jira or Linear acceptance criteria, runs against the Vercel preview and posts a report in the PR within minutes. An MCP server exposes 16 tools so Cursor, Claude Code or any MCP client can drive runs. Execution is cloud-native, in the vendor's own cloud, with no infrastructure to manage. On security, QA.tech advertises SOC 2, SSO and SAML, and states that customer data is never used to train models. Customer figures are the vendor's own estimates: Upsales is credited with replacing 320 hours of manual testing a month, Pricer with saving 390 hours a quarter, and the homepage projects 529% ROI, payback in three months and over 4 million USD saved across 36 months.
What it does
- Create tests from a plain-English description in a chat, with no script and no selector to maintain
- Run end-to-end journeys in a real browser, driven by intent rather than by a fixed sequence of steps
- Test every pull request against its preview deployment using the PR context, and report back before merge
- Test native iOS and Android apps on emulators, simulators and real devices
- Run scheduled exploratory sessions that look for defects outside the written scenarios
- Check WCAG accessibility on every run and catch visual regressions without pixel comparison
- Trigger and inspect runs from a coding agent through the MCP server, or from the CLI and the REST API
When to use QA.tech / When not to
A quick filter to help you decide if QA.tech is the right fit.
When to use QA.tech
- SaaS engineering teams that ship often and need regression cover to keep up: 18 of the 36 published use cases are tagged SaaS
- Product teams with no dedicated QA function, where PMs, designers and support staff have to write the tests themselves
- Teams whose scripted suites break on every UI refactor, and who spend more time fixing tests than writing them
- E-commerce teams checking checkout, cart, catalogue and account journeys on every release
- Engineering groups shipping AI-generated code from Cursor or Claude Code, who want runs triggered straight from the coding agent through MCP
When not to use QA.tech
- Buyers who need a published price or a self-service sign-up: all three tiers route to a sales conversation
- Teams after load or performance testing, since the platform covers functional, visual and accessibility checks only
- Developers looking for unit or isolated component tests, which the vendor's own Cypress comparison points elsewhere for
- Organisations that require self-hosted or on-premise execution, as runs happen exclusively in the vendor's cloud
- Teams needing low-level control over browser APIs or application state, which QA.tech's own comparison concedes is still Cypress territory
How to use QA.tech
A typical end-to-end flow, from setup to results.
- Book a 30-minute demo or ask for the free proof of concept, as there is no self-service sign-up
- Declare your project, your application and your environments (staging, preview, production)
- Let the agent crawl the application and build its feature graph, with no upfront configuration
- Hand over test credentials and test accounts, following the dedicated authentication documentation
- Open the network path where needed: allowlist the outbound IPs, set up an SSH tunnel, or clear Vercel preview protection and Cloudflare WAF and Turnstile
- Describe the journeys to check in plain English in the chat, adding business rules and edge cases, which the agents then keep between runs
- Organise the tests into test plans, with dependencies and shared steps
- Connect the CI/CD chain: the GitHub App for pull-request reviews, GitHub Actions, GitLab, Bitrise or Envoyer
- Trigger runs in parallel from the interface, the CLI, the API, or a coding agent through MCP
- Review the results (screenshots, logs, network activity, agent reasoning), track the issues, and route notifications to Slack, Microsoft Teams, Jira or Linear
Pros & Cons
Pros
- Test maintenance largely disappears: the suite is designed to survive refactors and redesigns, which is the vendor's central argument
- Non-developers can contribute, since PMs, designers and support staff write tests in plain English
- Fast authoring: the vendor puts a test at 2 to 5 minutes, against 30 to 120 minutes for the equivalent Cypress test
- No repository access is required, which lightens the security review, and the vendor advertises SOC 2, SSO and SAML plus a commitment never to train models on customer data
- Broad coverage on a single platform: web, native mobile, pull requests, exploratory runs, accessibility and visual regressions
- Fits the existing workflow through GitHub, GitLab, Vercel, Slack, Jira and MCP, with nothing to host or operate
- Authoring, chat and maintenance are included in every plan, only executions count against it, and test definitions are said to stay portable
Cons
- No public price and no self-service sign-up: the three tiers all lead to a sales conversation, and a demo or PoC is the only way in
- The site contradicts itself on certification: the footer and pricing page advertise SOC 2 Type 2 compliance, while the FAQ on that same pricing page says Type 1 certified with Type 2 in progress
- No hosting country or region is stated anywhere, and no DPA is published or mentioned; the trust centre that would hold those documents renders only in a browser
- Test data transits the vendor's cloud during every run, and no self-hosted option is offered
- The entry tier is tight: Starter caps at 1 user, 3 parallel runs and 2 environments, and keeps test data for 30 days only
- No load or performance testing, and no unit or isolated component testing
- The ROI and savings figures (529%, 4 million USD, 320 hours a month) are vendor estimates rather than audited measurements; the interface and support are English-only, and the privacy policy is generic and undated
Pricing & Plans
There is no permanently free plan and no public price. The three tiers on the pricing page, Starter, Growth and Enterprise, all carry the same Talk to us button, so every entry point runs through a sales conversation. No amount, currency or billing unit is published; the homepage structured data only records a minimum price of 0 USD together with the note Contact sales for pricing, which is not a usable starting point. The vendor does offer a free proof of concept covering two or three critical journeys, and states that test authoring, chat and maintenance are included in every plan, with only executions counted against it under a fair-use policy. The one price quoted on the site, 75 to 300+ USD a month for Cypress Cloud, refers to a competitor and not to QA.tech.
- 1 user
- 3 parallel runs
- standard model access
- 2 environments
- exploratory testing
- 30-day test data retention
- self-serve integrations and onboarding
- email support
- up to 5 users
- custom parallel runs and custom environment count
- multi-application
- mobile and pull-request testing
- coverage and run performance reports
- 90-day retention
- native integrations
- guided PoC
- custom user count
- unlimited parallel runs
- fine-tuned model option
- unlimited environments
- faster runs
- custom retention
- bespoke integrations
- dedicated support with an SLA
- All three tiers share the same Talk to us call to action
- which opens the demo booking form
- and none displays a price
Data, GDPR & hosting
A consolidated view of how QA.tech handles your data.
GDPR overview
GDPR is addressed explicitly: section 01 of the privacy policy is headed General Data Protection Regulation (GDPR), QA.tech declares itself a Data Controller there, and it lists six rights for European Economic Area residents, namely access, update or deletion, rectification, objection, restriction, portability and withdrawal of consent. Retention is described as lasting only for as long as is necessary for the purposes set out in that policy. The company is established in Sweden, inside the EU, so an Article 27 representative is not required and none is named. The weaker points: no Data Protection Officer is named, there is no dedicated privacy or GDPR address, requests go through the general contact form, and the page carries no effective date and no last-updated date. SOC 2 is claimed and sub-processors are published through a Vanta trust centre.
Who owns the data?
The only legal document readable on the site is the privacy policy, and it covers website visitors rather than test data: QA.tech declares itself a Data Controller of that information and lists four legal bases, namely performance of a contract, consent, legitimate interest and legal obligation. The main customer agreement is a PDF served by the Vanta trust centre and could not be read, so ownership of test cases, runs and results is not documented publicly. What the site does state is that customer data is never used to train models, that the agents never touch your source code, and that test data is kept 30 days on Starter, 90 days on Growth and a custom period on Enterprise.
Reuse rights
No clause covering reuse by the end user could be found. The privacy policy deals only with the data of website visitors, and the main agreement, which is the document that would set out what customers may do with their test definitions and results, is a PDF hosted on the Vanta trust centre that could not be read at collection time, so this point remains undocumented. The one relevant statement published on the pricing page is that there is no vendor lock-in and that test definitions stay portable and open. In practice users retrieve run artefacts, including screenshots, logs, network activity and the agent's reasoning at the point of failure, but no licence terms are stated for any of it.
Data retention & training
Hosting summary
QA.tech describes its execution model as cloud-native, stating that its agents run entirely in its own secure cloud, with no infrastructure for the customer to operate. Beyond that, no hosting country and no region is stated anywhere on the site. AWS is named for mobile emulators and simulators, and Amazon Device Farm for real devices, but neither is presented as the location of customer test data. Two technical signals should not be mistaken for hosting disclosures: the qa.tech domain resolves to 216.150.16.1, an anycast node on AS16509 (Amazon.com Inc.) geolocated in the United States, which describes the marketing site rather than the testing platform; and the trust centre is served from Vanta's European instance (app.eu.vanta.com), which says where the compliance tool runs, not where customer data sits. Sub-processors are said to be published at trust.qa.tech/subprocessors, but that page renders only in a browser and could not be read during collection. Anyone with data residency requirements will have to get this in writing from the vendor.
Things to keep in mind
Risks and trade-offs to weigh before adopting QA.tech.
- Certification to verify: the footer and the pricing page advertise SOC 2 Type 2 compliance, while the FAQ on that same pricing page says Type 1 certified with Type 2 in progress. It is a contradiction internal to the site, and it cannot be settled from the outside
- No DPA is published or explicitly offered, and no hosting country or region is named; the compliance documents sit in a trust centre that renders only in a browser, so any verification has to be done by hand
- The main customer agreement is a PDF served through that same trust centre and could not be read, so the contractual terms on test data ownership remain unverified
- The agents drive your environments from the vendor's cloud, so whatever appears on screen during a run, including test account content and production-like data, passes through it
- Test data is kept for only 30 days on the entry tier, and the privacy policy is generic with no effective date and no revision date, which is worth raising if you need an audit trail
- With no public price, no budget can be estimated and no offers compared without first investing time in a sales conversation
- Over-trust is the human risk here: a green run only proves that the journeys someone thought to describe still work. A team that stops using its own product by hand loses the instinct for the defects nobody wrote down
Setup & Integrations
Technical difficulty
Low to moderate. Nothing is installed and no repository access is granted: the agents run in the vendor's cloud and learn the application by crawling it, so a non-technical user can write tests in plain English straight away. An engineer is still needed for the plumbing, namely handling authentication and test accounts, allowlisting the platform's outbound IPs on a CDN, WAF or firewall, clearing Vercel preview protection or Cloudflare Turnstile, opening an SSH or remote tunnel to a local environment, and installing the GitHub App for pull-request reviews. The vendor claims a stable suite within days.
Deployment
Integrations
Supported languages
Behind QA.tech
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement QA.tech.
Frequently asked questions
Do you need to write code to create a test?
Does QA.tech plug into CI/CD and test pull requests?
Are native mobile apps covered?
How much does QA.tech cost?
Can you try the platform before committing?
How long before the tests are useful?
Is customer data used to train the models?
Where is the data hosted, and which certifications apply?
Should you pick QA.tech?
QA.tech makes a clear bet: replace scripted test suites with agents that pursue an intent, so a redesign no longer breaks the tests and QA stops being the exclusive business of JavaScript developers. Judged on the site, the engineering behind it is substantial, with a detailed documentation set, a REST API, a CLI, an MCP server exposing 16 tools, and CI/CD integrations covering GitHub, GitLab, Bitrise, Envoyer, Vercel and Cloudflare. Native mobile testing on emulators and real devices, pull-request testing against preview deployments and automatic accessibility checks place it beyond the narrower AI test-recorder category. The company is young: founded in 2023, backed by 5 million USD, a small Stockholm team, and two named customers, Upsales and Pricer. Its efficiency claims, 529% ROI, payback in three months and over 4 million USD saved across 36 months, are the vendor's own estimates, published without audit, and should be read as such. Several things need checking before a purchase. No price is published, all three tiers go through sales, and the only way to evaluate the product is a scoped free proof of concept. Nothing is said about where test data is hosted, no DPA is published, the trust centre that would hold the compliance documents renders only in a browser, and the site contradicts itself on SOC 2, claiming Type 2 compliance in the footer and Type 1 with Type 2 in progress in its own FAQ. The main customer agreement is a PDF that could not be read, so ownership of test data is not publicly documented. For a SaaS or e-commerce product team that ships fast and has no dedicated QA, QA.tech is worth a serious look, with those questions put to the vendor first.
- Choosing a selection results in a full page refresh.
- Opens in a new window.