Bugzy AI
Bugzy AI is an autonomous QA platform that generates, runs, self-heals and triages Playwright end-to-end tests on every pull request and deployment, then reports findings with screenshots and reproduction steps straight into Slack or Teams.
What is Bugzy AI?
Bugzy AI is an autonomous QA platform built for engineering teams whose testing capacity has fallen behind their release cadence. Instead of asking engineers to write and maintain scripts, it puts an AI agent in front of the product itself. Connect a repository, an issue tracker and a documentation source, hand it the application URL and a plain-English description of what the product does, and the agent browses the app, assembles a knowledge graph of how it is supposed to behave, and generates a full end-to-end suite. The vendor reports suites of forty or more test files produced within minutes.
The output is deliberately unremarkable, and that is the point: standard Playwright code, committed to the customer's own repository. Tests execute in real Chromium, Firefox and WebKit browsers, and they keep working with any Playwright-compatible runner if the subscription ever ends. Zero vendor lock-in is the central claim, set explicitly against the proprietary formats of the recorder-based tools Bugzy compares itself to.
Once the suite exists it runs on every pull request, push, deployment or cron schedule, reporting back as a GitHub check, a Slack or Teams message, or a ticket in Jira, Linear, Asana, ClickUp or Azure DevOps. Two capabilities carry most of the value. Self-healing rewrites selectors when the interface shifts, because the agent reasons about the intent of a test rather than matching a fragile locator. AI triage then sorts every failure into one of four buckets — genuine product bug, test needing adjustment, flaky run, or environment issue — so only actionable findings reach the team, each with screenshots, reproduction steps, severity and root-cause analysis. Teams that already own a Playwright suite can hand it over under a bring-your-own-tests model, and Bugzy will execute, monitor, triage and auto-fix it alongside the tests it writes.
Under the hood the agent runs on Anthropic's Claude API, and the conversational Slack and Teams assistant is built on the Claude Agent SDK. Coverage is web-first: the vendor states plainly that native mobile testing is out of scope. One caveat frames all of it — Bugzy has been folded into Felan, and every sign-up, documentation and sales path now leads to felan.ai.
What it does
- Generate complete end-to-end Playwright test suites from a product description and an application URL
- Run the suite automatically on every pull request, push, deployment or cron schedule
- Triage each failure into product bug, test to adjust, flaky run or environment problem
- Self-heal tests when the interface changes, without touching brittle selectors
- File bug reports with screenshots, reproduction steps, severity and root-cause analysis
- Take over an existing Playwright suite, then execute, monitor, triage and auto-fix it
- Test REST API endpoints alongside browser tests using the same reporter and infrastructure
When to use Bugzy AI / When not to
A quick filter to help you decide if Bugzy AI is the right fit.
When to use Bugzy AI
- Engineering teams whose developer-to-QA ratio has passed 8:1 and whose regression coverage no longer keeps up with releases
- Product teams with no dedicated QA engineer at all, who need automated coverage without waiting three to six months for a hire
- Teams already invested in Playwright, who want AI triage, self-healing and auto-fixing layered on top of the suite they own
- DevOps and platform engineers who want quality gates wired into GitHub Actions, GitLab CI or Jenkins and reported back into Slack or Teams
- Regulated organisations that need EU data residency, a published DPA and, on Enterprise terms, a self-hosted deployment
When not to use Bugzy AI
- Teams testing native iOS or Android apps: the vendor states outright that mobile testing is out of scope
- Anyone looking for visual or pixel-level regression testing, which the vendor itself points to specialised tools for
- Individuals and consumers: the terms restrict the service to legal entities acting in a professional capacity, and to users aged 18 or over
- Small teams on a tight budget, since the entry plan costs 250 euros a month with no free tier and no advertised trial
- Anyone who needs to sign up today, because the product has moved to Felan and every onboarding path now leaves the site
How to use Bugzy AI
A typical end-to-end flow, from setup to results.
- Connect the toolchain: a GitHub or GitLab repository, a Slack or Teams workspace, and an issue tracker. The vendor budgets five minutes for this step
- Provide the URL of the web application to be tested and describe the product and its key user flows in plain language
- Let the agent browse the application and read pull requests, tickets and documentation until it has built its context
- Review the first generated Playwright suite, which lands as standard code in the connected repository
- Configure event triggers so tests fire on push, pull request, PR comment or deployment, with filters per environment
- Add scheduled runs with cron expressions from the workspace settings for nightly regression passes
- Set up test data: typed environment variables for text, secrets, OAuth credentials and uploaded files
- Define named environments such as staging or production, with per-environment variable overrides
- Read the triage results where the team already works, in Slack, Teams or the issue tracker, and dispute any misclassification so the agent learns
- Set overage budgets with alerts at 50, 75 and 100 percent to cap monthly spend beyond the plan allowance
Pros & Cons
Pros
- Tests are standard Playwright files committed to your own repository, so they survive the subscription and carry no proprietary format
- No test code, selectors or maintenance scripts to write, which is the whole point of the product
- Triage separates real bugs from broken tests and flaky runs, so the team stops investigating every red build
- Setup is self-serve and fast, measured by the vendor in minutes rather than the weeks a managed QA service needs
- Coexists with an existing test suite instead of demanding a migration
- Pricing is published and self-serve, which is unusual in this category: several competitors the site itself lists show no prices at all
- Unusually thorough compliance material for a small publisher: EU-region hosting, a full DPA with named sub-processors, and a contractual no-training commitment
Cons
- The product has been absorbed into Felan, so there is no Bugzy sign-up, documentation or support path left on the site
- No free plan and no advertised free trial: the entry point is 250 euros a month
- No native mobile testing, a limitation the publisher states itself
- SOC 2 is described as a set of principles followed and a certification being worked toward, not one that has been obtained
- The ISO 27001 and SOC 2 certifications cited in the DPA belong to the hosting providers, not to the publisher
- Support is email only, in Bulgarian office hours, and there is no contact page: that URL now redirects off-site
- Monthly billing only in self-serve, with no pro-rata refund if you cancel mid-period, and any feedback you give becomes the vendor's property
Pricing & Plans
There is no permanent free plan and no free trial advertised on the pricing page. The lowest paid entry point is the Starter plan at EUR 250 per month, which includes 50 triages, 10 test case creations and up to 1,500 managed test runs, with overage billed at EUR 5 per additional triage and EUR 10 per additional test case. All prices are quoted excluding VAT, billed monthly in advance; annual billing is available only by arrangement with the sales team.
- 50 triages
- 10 test cases
- up to 1
- 500 managed test runs
- all integrations
- email support
- overage at EUR 5 per triage and EUR 10 per test case
- 100 triages
- 25 test cases
- up to 3
- 000 managed test runs
- priority email support
- advanced analytics
- overage at EUR 4 per triage and EUR 8 per test case
- 300 triages
- 50 test cases
- up to 8
- 000 managed test runs
- dedicated support
- custom reporting
- overage at EUR 3 per triage and EUR 6 per test case
- SSO/SAML
- dedicated account manager
- SLA guarantees and on-premise deployment
Data, GDPR & hosting
A consolidated view of how Bugzy AI handles your data.
GDPR overview
Implementation is concrete rather than declarative. The publisher is established in Bulgaria, so no Article 27 representative applies, and the privacy policy states compliance with the GDPR and the Bulgarian Personal Data Protection Act. Legal bases are set out per purpose under Article 6(1)(a), (b), (c) and (f). Seven data subject rights are listed with their articles, including portability in JSON or CSV, with a 30-day response commitment. A full Data Processing Agreement is published as an appendix to the terms, naming sub-processors and technical measures. Breach notification is 72 hours to the Bulgarian CPDP and 24 hours to the controller. Transfers to Anthropic and Stripe in the United States rely on Standard Contractual Clauses. A named data protection contact, Milko Slavov, is reachable at privacy@bugzy.ai.
Who owns the data?
Customers keep every right to the content they submit; Cloud Automation Solutions EOOD receives only a non-exclusive licence to process it in order to deliver the service. AI-generated output — test plans, test cases, bug reports — belongs to the customer outright, while the vendor retains its underlying models, algorithms and methodologies. Source code is explicitly out of bounds: access is read-only, and the terms state that the vendor does not access, modify or store it. One asymmetry deserves attention: any feedback a customer volunteers becomes the exclusive property of the vendor, irrevocably, with no compensation and no residual claim.
Reuse rights
Customers may reuse everything the platform produces without asking permission. Generated tests are committed as standard Playwright code to the customer's own repository and keep working with any Playwright-compatible runner after cancellation, and the AI-generated test plans, test cases and bug reports belong to the customer. On the vendor's side, data is processed only to run the service: generating and executing tests, delivering results, managing integrations, support, security monitoring and legal compliance. Customer data is never used to train AI models — an undertaking repeated in the terms, the privacy policy and the DPA, and enforced contractually against Anthropic. Selling data, unrelated marketing and disclosure of confidential business information are all excluded. Customers are in turn barred from using the service for employee evaluation, HR decisions, biometric identification or law enforcement.
Data retention & training
Hosting summary
Data is stored primarily inside the European Union. The privacy policy names four EU-region providers: Vercel for web hosting and edge computing, Supabase for the database and authentication, Google Cloud Platform for file storage and compute, and PostHog for product analytics. The jurisdiction governing the whole arrangement is Bulgarian and, above it, European Union law, with the Bulgarian Commission for Personal Data Protection as supervisory authority. One documented exception applies: test generation sends a minimised volume of data to Anthropic's Claude API in the United States, and payments involve Stripe, also in the United States. Both transfers rely on Standard Contractual Clauses backed by a Data Processing Agreement. Test artefacts such as screenshots, traces and logs sit in per-team isolated storage. Encryption is TLS 1.3 or above in transit and AES-256 at rest. Enterprise customers are offered data residency options and on-premise deployment. Note that the site's resolved IP address is an anycast CDN node and says nothing about where customer data actually lives.
Things to keep in mind
Risks and trade-offs to weigh before adopting Bugzy AI.
- Product continuity is the first risk: Bugzy has been folded into Felan, so any evaluation based on this site describes a service you can no longer subscribe to here, and every claim dates from March 2026
- Trusting the triage too far is the classic failure mode of this kind of tool: a green suite proves the generated tests passed, not that the product is correct, and the vendor itself insists that AI output must be reviewed by a qualified human
- Skill atrophy is a real cost: a team that never writes a test again gradually loses the judgement needed to tell a valuable test from a shallow one, and to notice what the agent never thought to cover
- Access breadth deserves scrutiny before signing: the agent reads repositories, tickets, documentation and team conversations, so the customer must confirm it has the rights to grant that access and avoid exposing special category data
- Data leaves the European Union for AI processing at Anthropic in the United States under Standard Contractual Clauses, which is disclosed but should be assessed against your own transfer policy
- Security assurance is weaker than it first appears: SOC 2 is an ambition, not a certification, and the ISO 27001 and SOC 2 mentions in the DPA describe the hosting providers rather than the publisher
- Two contractual asymmetries are easy to miss: any feedback you provide becomes the vendor's exclusive property without compensation, and cancelling mid-period brings no refund
Setup & Integrations
Technical difficulty
Low for an engineering audience, and deliberately so. Connecting the repository, chat workspace and issue tracker is budgeted at five minutes; after that you supply an application URL and a product description, and the agent does the rest. No test code, no selectors and no infrastructure to stand up. Deeper configuration — event triggers, named environments, typed test data, cron schedules, OAuth credentials for the application under test — is optional and comes later. The genuine prerequisite is administrative rather than technical: you must hold the rights to grant read access to the systems being connected.
Deployment
Integrations
Behind Bugzy AI
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Bugzy AI.
Frequently asked questions
What exactly does Bugzy AI do?
Do I have to write any test code?
What technology are the tests built on, and can I keep them?
Can it work with the test suite I already have?
How much does it cost, and is there a free plan?
Where is my data hosted, and is it used to train AI models?
Does it read my source code?
Is Bugzy AI certified to SOC 2 or ISO 27001?
Is the product still available?
Should you pick Bugzy AI?
Bugzy AI is one of the more technically honest offers in the autonomous QA category. Its central promise is unusual in that it argues against its own stickiness: the tests it writes are ordinary Playwright files, committed to your repository, usable without the product. Everything else follows from that position — triage that sorts real bugs from broken tests, self-healing tied to test intent rather than selectors, and a bring-your-own-tests mode that improves a suite you already own instead of replacing it. The compliance material is unusually complete for a two-person publisher: EU-region hosting, a full DPA with named sub-processors and technical measures, and a no-training commitment repeated across three contractual documents. Prices are published, which several competitors the site itself lists do not do.
Two reservations matter more than the rest. The first is status: the product has been absorbed into Felan, and every sign-up, documentation and sales path now leaves the site. What remains is an intact record of what Bugzy was, not a service you can subscribe to here. The second is security posture: SOC 2 is presented as a set of principles followed and a certification in progress, and the ISO 27001 and SOC 2 badges appearing in the DPA belong to Google Cloud, Vercel and Supabase rather than to the publisher. Read those two together before treating any claim on the site as current.
Within its stated boundaries — web applications only, business customers only, entry at 250 euros a month — the design remains a sound reference for what autonomous QA should look like. Anyone acting on it today should start at felan.ai and confirm which capabilities carried over.
- Choosing a selection results in a full page refresh.
- Opens in a new window.