midscene
Midscene is a vision-based GUI agent and testing framework for end-to-end tests of web, mobile, and desktop apps. Describe interface actions and expected results in natural language, then use it to interact with screens and verify what users see.
Pick a small issue you can reproduce, then fix it with your AI coding agent.
Share on XLicense: MIT
Overview
Midscene is a vision-based GUI agent and testing framework for end-to-end tests of web, mobile, and desktop apps. Developers describe interface actions and expected results in natural language, and the agent uses screenshots to locate controls, interact with them, and check visible outcomes. Tests can be organized as YAML cases with Midscene Test or run through Agent APIs in existing Playwright tests. Generated HTML reports show screenshots, actions, and assertion results.
Key features
- Describe UI actions, waits, and assertions in natural language.
- Locate controls from screenshots and click, type, or scroll without maintaining selectors.
- Write YAML test cases with Midscene Test or use Agent APIs with Playwright.
- Inspect screenshots, actions, and assertion results in generated HTML reports.
Best for
Developers who want to test user-visible behavior across web, mobile, or desktop interfaces. Choose it when tests should describe what a user sees and does rather than rely on maintained selectors.
- Upstream
- web-infra-dev/midscene
- Fork on GitHub
- Guo-astro/midscene
- Upstream stars
- 16k
- Category
- Developer tools and infrastructure
- Language
- TypeScript
- License
- MIT
- Forked
- 2026-10-13
- Sync status
- In syncLast synced 2026-11-08
More in Developer tools and infrastructure
An open-source TypeScript engine for calculating Zi Wei Dou Shu charts and consulting traditional references. It helps developers build charting and interpretation tools.
Forked 2026-10-18Last synced 2026-11-08License: MITDeveloper tools and infrastructureGitHub
ImHex is a hex editor for inspecting and editing binary files. Reverse engineers, programmers, and forensic users can search and patch bytes or describe file structures with its pattern language.
Forked 2026-10-17Last synced 2026-11-08License: GPL-2.0Developer tools and infrastructureGitHub
Numa is a portable DNS resolver for local development and privacy. It blocks ads, gives local services names like frontend.numa, and runs without depending on a cloud account.
Forked 2026-10-17Last synced 2026-11-08License: MITDeveloper tools and infrastructureGitHub
Handwired turns ordinary JavaScript logic into differentiable neural-network-style calculations, so programs can run as matrix operations in JavaScript or WebGPU. It is for experimenting with GPU execution and optimization of code written in a traditional style.
Forked 2026-10-16Last synced 2026-11-08License: MITDeveloper tools and infrastructureGitHub