testosaurusdocs
Browse docs
On this page

Privacy Policy

Effective 2026-10-17 · Testosaurus is operated by Goose Data S.R.L., Strada Liviu Rebreanu nr. 10, Corp C, Camera 1, Bistriţa, jud. Bistriţa-Năsăud, Romania · Trade Registry J2024049768006 · CUI 51047393 ("Testosaurus", "we") · Contact: privacy@testosaurus.dev

1. Our two roles

For your account data, Testosaurus is the controller: we decide why and how it is processed. For the feedback data your application's users submit through widgets you embed, you (the Testosaurus customer) are the controller and we are the processor, acting on your instructions. If you are an end-user who sent feedback through a widget on someone else's site, that site's operator is responsible for your data — their privacy notice applies, and this policy describes how we handle it on their behalf.

2. What we collect

  • Account data: email address, name, a password hash (authentication is handled by Better Auth; we never store plaintext passwords), whether the address is confirmed, plan and workspace membership, and optional invite, referral, beta or offer codes you use, with their status.
  • Billing data: Stripe customer and subscription identifiers, the prepaid balance ledger, your automatic-refill settings, usage counters (accepted submissions, storage, AI usage) and the billing profile you give us for invoices: name, address and, for a business, its tax identifier. Card details are entered in Stripe and never reach our systems.
  • Feedback data (submitted via the widget on your site): the report text, the page URL and title, user agent, viewport, language and recent console warnings and errors, cleaned of secrets as described in section 8, and optionally a screenshot and a session replay. Screenshots and replays are opt-in and require end-user consent in the widget.
  • Coding-run data: when you start a coding run, the report text and context, the repository files the agent reads and changes inside your own GitHub Actions runner, the agent's conversation with the model, test output and the resulting draft pull request.
  • Security data: IP addresses (from the trustworthy cf-connecting-ip header only) for rate limiting and abuse protection, Cloudflare Turnstile verification tokens, and the hostname a report was sent from. When you connect a coding agent, we also keep the approximate city and country that Cloudflare derives from the IP address of the agent that starts the request, of the browser that approves it and of the agent that collects the key. We show them on the approval page and in the confirmation email we send you, to protect you against approval phishing, and keep them on our servers for at most 7 days.
  • Operational data: queue and delivery logs (for example, analysis failures and webhook delivery results), email-sending counters and audit records of access-key activity.

3. What we do with it

  • Operate the service: store, show, filter, group and export your feedback.
  • AI features you enable: report text, context and, where the selected model supports vision, the screenshot are sent through the Vercel AI Gateway to the model provider the project owner selected from the model catalog, to write the structured analysis (title, category, severity, summary, steps), to decide whether two reports describe one problem (issue matching) and, for coding runs, to let the agent investigate and change code. We never use your data to train models; model requests go through Vercel AI Gateway to the provider you select, under that provider's API terms.
  • Coding runs: when you connect GitHub with the repo and workflow permissions and start a run, we install and dispatch a GitHub Actions workflow in your repository and open a draft pull request. Repository content is read inside your own GitHub Actions runner; the agent's conversation transits the AI Gateway to the model you selected. Testosaurus does not keep a copy of your repository; the model's responses are held only while a run is active.
  • Notify you: email alerts on high-severity reports, account and billing notices (Cloudflare Email) and webhook deliveries to URLs you configure.
  • Integrations: if you connect GitHub or Linear, issue payloads are sent to those services when you raise a report.
  • Billing and accounting: Stripe processes payments. StartCo issues our invoices and registers them with the Romanian tax authority, so it receives your billing name, email, address and tax identifier.
  • Abuse protection: Turnstile verification, rate limits, and content filters run on the public submission path.
  • Operator access: the people who run Testosaurus can see account, billing and usage state and the reports needed to answer a support request, and use audited tools such as confirming an email address or extending a beta grant.
  • Aggregate statistics: Cloudflare Web Analytics gives us cookieless, aggregate page-view statistics for testosaurus.dev; it does not track you across sites.

Contract performance for account and service operation (Art. 6(1)(b)); legal obligation for invoices and accounting records (Art. 6(1)(c)); legitimate interests for security, abuse prevention, aggregate analytics and service improvement (Art. 6(1)(f)); consent, obtained by the widget, for screenshots and session replays (Art. 6(1)(a)). End-user feedback is processed under the site operator's instructions — their lawful basis covers the submission.

5. Processors, sub-processors and recipients

  • Cloudflare — hosting and data storage (Workers, D1, R2, Queues), email delivery, Turnstile bot protection and Web Analytics.
  • Stripe — payments and the customer portal. Stripe receives your billing email, billing name and address, and payment details.
  • Vercel AI Gateway and the model providers you select — model requests for analysis, issue matching and coding runs are routed through the Vercel AI Gateway to the provider whose model the project owner chose from the model catalog, under that provider's API terms.
  • StartCo (Romania) — invoicing. It receives the buyer's name, email, address and tax identifier and the invoice lines of each payment.
  • ANAF RO e-Factura — the Romanian tax authority's e-invoicing system receives invoices as a recipient under a legal obligation (GDPR Art. 6(1)(c)).
  • GitHub / Linear — only when you connect them: OAuth access, issue payloads you raise, and for coding runs the GitHub Actions workflow, branch and draft pull request in your repository, and the state of pull requests linked to fixes in your connected repository, read from time to time for up to 30 days until they merge or close.

6. Where data lives and how long

Data is stored in Cloudflare D1 (metadata and feedback) and R2 (screenshots and replays) with edge locations worldwide.

  • Reports and attachments: reports, screenshots and replays are purged automatically 90 days after they were created on Free, 365 days on Pro and 3,650 days on Enterprise (the plan in force when the purge runs). Uploads not attached to a report are deleted 48 hours after upload.
  • Coding runs: summaries and test results are kept with their report until it is purged or deleted. Coding-run model transcripts are blanked when a run settles.
  • Logs: rate-limit and email-sending records are deleted after 7 days. Connect requests, with the approximate locations above, are deleted at most 7 days after they are created.
  • Accounting records: we keep accounting records — invoices, payment and e-Factura records, and the billing profile used on an issued invoice — for 10 years, the period the Romanian accounting rules require.
  • Everything else: account data is kept while your account exists.
  • Backups: deleted data can remain in encrypted backups for up to 30 days before the backups expire.

Deleting a project removes its reports, screenshots and replays, the issues they were grouped into with their AI summaries and developer guidance, tracker link records and the saved text of coding runs; the stored files are deleted and retried until they are gone. Deleting your account cancels your subscription, ends automatic refills, removes your projects as above, your integrations (GitHub and Linear access is revoked), team memberships and invitations, and setup progress records. Unused purchased credit is forfeited when you confirm deletion; email support@testosaurus.dev before deleting to ask for a refund. Billing and invoicing records are kept as the law requires, and the email addresses stored on offer, referral and beta records are blanked.

7. Your rights

You can download a project as JSON at any time from settings → data → Export. The export has the project's settings, its reports (text, context and analysis) and a manifest of its screenshots and replays, not the media files themselves, which stay viewable in the dashboard while they are stored. Invoices are listed under billing & usage. You can delete a project or your entire account yourself. On top of that, EU residents have the usual GDPR rights: access, rectification, erasure, restriction, portability, and objection — write to privacy@testosaurus.dev. End-users should contact the site operator they submitted feedback to; we support operators in fulfilling such requests through export and deletion tools. You may also lodge a complaint with your data protection authority.

8. Cookies, local storage and capture

The dashboard uses a session cookie for sign-in. During sign-up we also set a cookie for one hour, so that the confirmation link signs in only the browser you signed up in. We use no advertising cookies and no cross-site trackers, and nothing on the marketing site profiles you.

The widget uses your browser's storage on the site where it runs, only for what it needs to work:

  • testosaurus:pending (local storage) — reports waiting to be sent when you are back online, at most five.
  • testosaurus:screenshot-consent, testosaurus:replay-consent and testosaurus:active-consent (local storage) — your capture choices.
  • testosaurus:welcome-seen (local storage) and testosaurus:welcome-ducked (session storage) — whether you answered or dismissed the welcome bubble.

The dashboard and the documentation keep interface preferences (the selected project, list refresh, theme) and pending invitation or offer codes in local or session storage.

Redaction and masking. Before a report leaves the browser, the widget removes secrets from the page address (values of parameters such as token, code or password, fragments and credentials) and replaces tokens, credentials and email addresses found in the address, title and console output with [redacted]; the server applies the same rules before anything is stored or passed to AI. Typed values are masked in screenshots and replays, and site operators can mask or block other elements with data-testosaurus-mask and data-testosaurus-block. Redaction is pattern based and cannot recognise every secret.

9. Changes

Material changes to this policy will be announced in advance (changelog and, for account holders, email). The effective date above always reflects the current version.

Something unclear?Let us know