AI vision test generation
Describe a user flow in plain English and let the vision agent click, type, and read the screen before writing the test. The homepage says this avoids selectors, DOM access, and setup work.
TestDriver is an AI-powered testing platform for automating desktop and application workflows with pixel-based vision rather than selectors or DOM access. The homepage positions it as a way to make end-to-end testing faster, more reliable, and easier to maintain.
It is aimed at teams that need to test web apps, browser extensions, editor extensions, and desktop applications on real Windows, macOS, and Linux environments. The product combines AI-driven test creation, cached element matching, visual assertions, and CI-friendly execution in a single workflow.
The pricing page shows a free Cloud plan, paid Cloud plans, and an enterprise self-hosted option, with higher tiers adding more parallel tests, more included minutes, recordings, analytics, private support, and on the enterprise tier, custom implementation and environment control.
In practice, TestDriver is built for teams that want to describe flows in natural language, run them as JavaScript tests, and inspect run artifacts when something breaks. The site also highlights support for third-party web apps and rich UI content such as file uploads, PDFs, and embedded media.
Describe a user flow in plain English and let the vision agent click, type, and read the screen before writing the test. The homepage says this avoids selectors, DOM access, and setup work.
Tests are cached by element fingerprint so repeated runs can replay quickly and deterministically when the UI has not changed. The site says this reduces token use and keeps passing tests fast.
When an element changes location or appearance, TestDriver re-runs the AI to find it and updates the cache instead of forcing a rewrite. This is presented as a way to keep tests passing through UI changes.
TestDriver describes its tests as plain Vitest that run on real VMs across web, desktop, and extensions. That lets teams use a familiar test format inside their existing CI pipeline.
The product can validate outcomes with AI-backed assertions and return a boolean result with reasoning. The homepage notes that assertions are cached as well.
Debugging output includes video replay, network logs, request details, CPU and memory data, and a step-by-step action log. These views are meant to make failing runs easier to inspect.
Use TestDriver to describe a business flow in plain English and turn it into a runnable test that clicks, types, and validates the screen state without relying on selectors.
Run the same test format in CI so teams can verify app behavior on every commit or pull request and compare local and pipeline results.
Test browser extensions, editor extensions, and native desktop apps in their real environments, including interfaces that are difficult for selector-based tools to cover.
Inspect video replays, request logs, action history, and CPU or memory data to understand why a run failed and where the interaction diverged.
Use the cached element model and AI re-matching to keep tests stable when layouts shift, text moves, or UI components are redesigned.
TestDriver is designed to write tests from plain-English flow descriptions with its AI vision agent, then run those tests through a familiar JavaScript API. The homepage also shows an example setup command, `npx testdriverai init`, and the pricing page indicates a free Cloud tier and paid plans for more usage and platforms.
The homepage says TestDriver supports web apps, Chrome extensions, VS Code extensions, Windows apps, macOS apps, and Linux apps. The pricing page also lists mobile targets, but the enterprise self-hosted plan marks Android and iOS as coming soon.
The pricing page shows a free Cloud plan, a Pro Cloud plan, a Team Cloud plan, and an Enterprise self-hosted option with contact sales pricing. The page also notes different included minutes, parallel test limits, and support levels by plan.
TestDriver records test runs and can surface video replays, network logs, request details, CPU and memory data, and an action log. The homepage presents these as debugging and validation tools for understanding failures and assertions.
The source does not document a full onboarding guide, but it shows the product can be initialized with `npx testdriverai init` and then used with AI vision-based test generation and Vitest-style tests.
流量数据仅供参考。
codesnippets.ai
Code Snippets AI 让开发者结合代码库和安全代码片段库与 AI 对话,构建功能、修复问题并跨桌面与浏览器工作流共享代码。
browseros.com
BrowserOS 是一款开源 AI 浏览器,可通过本地代理根据自然语言提示自动执行网页任务,支持 macOS、Windows、Linux、Chrome 扩展及基于 MCP 的应用集成。
blopai.com
编写、运行、聚类并修复浏览器测试的 QA 代理
comate.baidu.com
百度 AI 编程助手,支持代码、单元测试、注释生成与问答,覆盖 100 多种语言和多款 IDE
tensorflow.org
开源机器学习平台,支持在桌面、移动端、Web、云端和边缘环境构建、训练与部署模型。
agents.craft.do
Craft Agents 是一款开源桌面应用,连接模型、API、MCP 服务器、本地文件和内置浏览器,助你一站式规划、执行和审查工作。