This page lists what you need to use Qure and where it stops. Limits marked as current are set by the service and can change. The app shows the values that apply to you.

​
Principal features

  • Qure is a desktop application for creating and maintaining end-to-end tests of web applications with an AI agent. It works in your own test repository.
  • Open a local folder, create an empty project, or clone a repository from GitHub or GitLab over HTTPS.
  • Attach your application’s source code as context for the agent, read-only or editable.
  • Record interactions in the built-in browser, with pause and resume, comments, checks and voice notes.
  • The agent reads your repository, writes and edits tests, runs them and repairs failures. It does not commit or open pull requests on its own. Agent actions need your approval, either one by one or automatically by action type.
  • Generated tests are ordinary framework code in your repository. They keep running without Qure.
  • Run tests and review results with per-step screenshots and traces. Playwright tests also get a built-in trace viewer.
  • Write manual test cases in Markdown and let the agent execute them in the built-in browser.
  • View CI runs, logs and artifacts for GitHub Actions and GitLab CI, and rerun or cancel them.
Qure does not generate unit or integration tests, run accessibility or security audits, or test across device and viewport matrices.

​
Operating systems

  • Windows on ARM is not available.
  • No RPM package is published. The .deb package needs GTK 3, NSS, libnotify, libxss, libxtst, xdg-utils, desktop-file-utils, AT-SPI and libuuid.
  • On Linux, install Qure in a path without spaces. The default /opt/Qure works.
  • Running tests inside WSL is not supported.
  • macOS builds are signed with a JetBrains Developer ID and notarized by Apple. Windows builds are signed by JetBrains.
  • Qure checks for updates at startup and then every hour, and downloads them automatically unless you turn this off in Settings. On macOS, automatic download works only when the app is in the Applications folder.
  • The service can require a minimum app version. Update older versions before you use them.

​
Browsers and test frameworks

  • The built-in browser, used for recording, agent exploration and manual test runs, is based on Chromium. It does not reproduce Safari- or Firefox-specific behaviour.
  • Your test suite can still target Firefox or WebKit through your framework’s configuration. Qure runs these tests without visual replay.
  • Qure does not bundle the browsers your tests use. It uses the browsers your framework installed, and can download Chrome for Testing when a specific version is needed.
Generated tests are written in JavaScript, TypeScript, Python or Java. Tested minimum versions: Node.js 22.18, Playwright 1.62, Cypress 14.2.1, Python 3.11 and Robot Framework 7.0. Older versions work on a best-effort basis.
  • You can test internal applications that are not on the public internet, as long as your machine can reach them. Corporate SSO policies that allow only an approved external browser can block recording and signed-in browsing.
  • The built-in browser accepts self-signed or invalid TLS certificates for the application under test and marks the page as not secure.

​
Requirements and setup

​
Account

  • You need a Qure account. Sign in with Google or GitHub. Sign-in opens in your system browser. Enterprise SSO is not supported.
  • Accept the Terms of Service before you open a project.
  • Accounts belong to an organization with Admin and Member roles. Invitations expire after 7 days.
  • The AI agent requires an active subscription or trial with available usage credit.

​
Network

Qure needs an internet connection. Sign-in and the AI agent run through Qure’s cloud service, and there is no offline mode. A short network drop does not sign you out. Allow outbound HTTPS to:
  • api.quretests.com, the Qure service;
  • download.jetbrains.com, for updates;
  • eu.i.posthog.com, for product analytics;
  • api.openai.com, for voice input only;
  • the hosts of the integrations you connect;
  • googlechromelabs.github.io, github.com and gitlab.com, if you let Qure download Chrome for Testing or the GitHub and GitLab CLIs.
Qure has no proxy setting of its own. Commands it runs for you inherit HTTP_PROXY and HTTPS_PROXY from your environment.

​
Local tools

  • Git must be installed. Qure detects it and can offer to install it.
  • Install your project’s own runtime and test tooling yourself, for example Node.js with your test framework, a Python environment, or a JDK with Maven or Gradle. Qure detects them but does not install them. Docker is not required.
  • CI features need the GitHub CLI (gh) or GitLab CLI (glab 1.53.0 or later), signed in with your own credentials. Qure can install them. GitLab instances older than 17.0 show fewer pipeline details.

​
Permissions and credentials

  • On macOS, Qure asks for microphone access only when you use voice input. It does not need Screen Recording or Accessibility permissions.
  • Credentials are stored in the operating system’s keychain or keyring. On Linux without a keyring, the sign-in token is stored in a file only your user can read, and integration credentials last for the current session only.

​
Integrations

  • Cloning from Bitbucket, Azure DevOps, Gitea or Codeberg is not supported.
  • Jenkins, TeamCity, CircleCI, Bitbucket Pipelines and other CI systems are not integrated.
  • Jira Data Center and Jira Server need a custom MCP connection.
  • Test results are not written back to test management tools automatically.
  • Qure chooses the AI models and calls the providers (Anthropic and OpenAI) from its cloud service. You cannot use your own API keys.

​
Regional notes

  • Qure’s cloud service runs on AWS in the EU (Ireland).
  • Prompts and context are processed by the AI providers Anthropic and OpenAI in their own processing locations. Qure does not guarantee EU-only processing.
  • Voice input connects from your computer directly to OpenAI, so it may not work where OpenAI is unavailable.
  • The Qure interface is available in English only. The applications you test can use any language.

​
Technical limitations

​
Usage and concurrency

  • AI usage is metered against your organization’s credit. When the credit runs out, agent runs stop until the next period or a top-up.
  • Trial organizations can run up to 3 agent runs at the same time.
  • Up to 5 manual test runs can run in parallel per user (3 by default). Further runs wait in a queue.

​
Size limits (current)

  • Chat attachments: up to 20 files per message, about 5 MB per file and 30 MB in total. You can attach text, Markdown, HTML, CSS, JavaScript, TypeScript, JSON, Python, PDF and common image formats (JPG, PNG, GIF, WebP, HEIC). Excel and CSV files are not accepted. Images are downscaled.
  • Uploaded test results: up to 1 GB per archive (5 GB unpacked, 100,000 files), 500 MB per attachment, and 100 GB per project in any 30 days.
  • Imported skills: up to 256 files of 1 MB each.
  • Up to 200 stored chats per repository by default.
  • Test commands that Qure does not manage time out after 120 seconds by default.
  • Long conversations are summarized automatically to fit the model’s context, so earlier details can be condensed.

​
Functional limits

  • Recording does not yet follow flows that move to a second tab or window.
  • Recorded steps cannot be edited, reordered or deleted one by one.
  • Manual test cases run only in the desktop app, not in CI.
  • A project has one test repository. You can attach your application’s source in addition.
  • Qure has no secrets vault. Use dedicated test accounts with non-production credentials.

​
Data handling

  • To do its work, the agent sends content to Qure’s cloud service and from there to the AI providers. This includes file contents it reads, command output, page content and screenshots of the application under test, recorded interactions, attachments and your messages.
  • The agent does not read credential files such as .env. Test-account passwords are entered locally and are not shown to the model.
  • Password, one-time code and card fields captured while recording are redacted in stored history. During an active session their values can still reach the AI provider.
  • AI providers can cache prompts for a limited time: currently up to 1 hour for Anthropic and 24 hours for OpenAI.
  • Uploaded test-result attachments are deleted after 30 days. Structured test results are kept for 365 days.
  • Qure collects usage analytics, without session recording, and crash reports.