cypress-mcp
Exposes Cypress as AI-usable tools for running tests, managing spec files, taking accessibility snapshots, and automating a browser via Playwright-core.
README
cypress-mcp
MCP server for Cypress — let AI agents run tests, manage specs, snapshot pages, and automate browsers.
What is this?
cypress-mcp is a Model Context Protocol server that exposes Cypress as a set of AI-usable tools. Connect it to Claude, Cursor, VS Code, or any MCP client and let your AI assistant:
- Run your Cypress tests and get structured pass/fail results
- Read and write spec files — generate new tests from natural language
- Snapshot any page via the ARIA accessibility tree — no screenshots needed
- Automate a browser — navigate, click, type, evaluate JavaScript
Think of it as the Cypress equivalent of @playwright/mcp.
How it works
Your AI (Claude / Cursor / VS Code)
│
│ MCP protocol (stdio or HTTP/SSE)
▼
cypress-mcp server
│
├── Spec tools ──────────► Filesystem (list / read / write .cy.ts files)
├── Runner tools ─────────► Cypress CLI (run tests, open GUI, get results)
└── Browser tools ────────► Playwright-core (navigate, snapshot, screenshot)
Browser automation uses a separate Playwright-core instance — not the Cypress browser — keeping automation and testing concerns cleanly separated.
Quick Start
1. Install browsers (one-time)
npx playwright install chromium
2. Connect to your MCP client
Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"cypress": {
"command": "npx",
"args": ["cypress-mcp", "--project", "/path/to/your/app"]
}
}
}
VS Code / Cursor (HTTP/SSE mode):
npx cypress-mcp --port 8932 --project /path/to/your/app
Then point your MCP client at http://localhost:8932/sse.
3. Start asking your AI
> List the spec files in my project
> Navigate to http://localhost:3000/login and take a snapshot
> Write a Cypress test for the login form based on what you see
> Run the login spec and tell me what failed
Tools
Spec file tools
| Tool | Description |
|---|---|
cypress_list_specs |
List all spec files matching Cypress patterns |
cypress_read_spec |
Read a spec file's contents |
cypress_write_spec |
Create or overwrite a spec file |
cypress_get_config |
Read cypress.config.ts/js |
Test runner tools
| Tool | Description |
|---|---|
cypress_run |
Run tests — returns pass/fail counts and error details |
cypress_open |
Open Cypress GUI |
cypress_get_last_run |
Read results from the last run |
Browser automation tools
| Tool | Description |
|---|---|
cypress_navigate |
Navigate to a URL |
cypress_snapshot |
Get ARIA accessibility tree of the page |
cypress_screenshot |
Take a screenshot (full page or element) |
cypress_click |
Click an element by selector or text |
cypress_type |
Type into an input field |
cypress_evaluate |
Run JavaScript and return the result |
cypress_get_url |
Get the current URL |
cypress_wait_for |
Wait for an element or text to appear |
cypress_close_browser |
Close the automation browser |
Example workflow
Here's what happens when you ask an AI to write tests for your app:
1. Snapshot the page — cypress_navigate + cypress_snapshot
Accessibility Snapshot:
- heading "My Todo List" [level=1]
- textbox "New todo"
- button "Add"
- list "Todo list"
- listitem: No todos yet. Add one above!
2. AI writes a spec — cypress_write_spec
describe('Todo App', () => {
it('adds a new todo', () => {
cy.visit('http://localhost:4000');
cy.get('[aria-label="New todo"]').type('Buy groceries');
cy.get('button').contains('Add').click();
cy.get('#todo-list').should('contain', 'Buy groceries');
});
});
3. Run the tests — cypress_run
✅ Cypress Run PASSED
Tests: 10
Passed: 10
Failed: 0
Duration: 3.21s
CLI Options
cypress-mcp [options]
--project <path> Cypress project root (default: current directory)
--port <number> Start HTTP/SSE server instead of stdio
--browser <name> Browser for automation: chromium | firefox | webkit (default: chromium)
--headed Show the automation browser window
--help Show help
--version Show version
Running Locally
git clone https://github.com/yashpreetbathla/cypress-mcp.git
cd cypress-mcp
npm install
npx playwright install chromium
npm run build
npm test
Development mode
npm run dev # TypeScript watch mode
npm test # Run unit tests (Vitest)
npm run test:watch # Watch mode
Requirements
| Requirement | Version |
|---|---|
| Node.js | >= 18 |
| Cypress (in your project) | >= 12 |
| Chromium (for browser tools) | via npx playwright install chromium |
Cypress is a peer dependency — install it in your project, not globally. The MCP server picks it up automatically from node_modules/.bin/cypress.
Architecture
cypress-mcp/
├── src/
│ ├── index.ts CLI entry point
│ ├── server.ts MCP server (stdio + HTTP/SSE transports)
│ ├── context.ts Playwright-core browser context singleton
│ ├── toolHandler.ts Central tool dispatcher
│ ├── tools/
│ │ ├── specs.ts Spec file management (filesystem)
│ │ ├── runner.ts Cypress test execution (CLI)
│ │ └── browser.ts Browser automation (Playwright-core)
│ ├── types.ts TypeScript interfaces
│ └── version.ts Version constant
└── tests/
└── unit/
├── specs.test.ts 13 tests for file tools
├── tools.test.ts 8 tests for tool definitions
└── toolHandler.test.ts 4 tests for dispatch logic
Why Playwright-core for browser automation?
Cypress is a testing framework — its browser isn't accessible as a real-time automation API. Rather than try to hack into the Cypress browser, cypress-mcp runs a lightweight separate Playwright-core instance for automation tasks (navigate, snapshot, screenshot). The Cypress browser is only used when running actual tests via cypress_run.
Comparison with @playwright/mcp
| Feature | cypress-mcp |
@playwright/mcp |
|---|---|---|
| Browser automation | ✅ (via Playwright-core) | ✅ (native) |
| Accessibility snapshot | ✅ ARIA tree | ✅ ARIA tree |
| Run existing tests | ✅ Cypress tests | ❌ |
| Write test files | ✅ .cy.ts / .cy.js |
❌ |
| Test results parsing | ✅ structured JSON | ❌ |
| Vision / coordinate tools | ❌ (coming) | ✅ |
| Best for | Teams using Cypress for E2E | General browser automation |
Contributing
Issues and PRs welcome. Please open an issue before starting significant work.
npm test # must pass
npm run build # must compile cleanly
License
MIT © Yashpreet Bathla
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。