---

title: ClaudeTest emoji: 🧪 colorFrom: blue colorTo: purple sdk: docker pinned: false app_port: 7860

ClaudeTest

Enterprise-grade Playwright (web) and Appium (mobile) test automation plugin for Claude.

Connect your GitHub repository and let Claude generate missing tests, diagnose CI failures, and open fix PRs — all from a single conversation.

---

Features

• Test generation — Claude scans your codebase and writes Playwright (web) or Appium (mobile) tests for untested flows
• CI diagnosis — paste a failed GitHub Actions run ID and get a plain-English root cause + fix suggestion
• Auto PR — generated tests are committed to a new branch and opened as a pull request
• Cross-platform mobile — Appium tests cover iOS and Android from the same test file
• Self-hostable — run inside your own VPC with Docker; no code leaves your infrastructure

---

How it works

You → Claude → ClaudeTest MCP server → GitHub API / GitHub Actions
                                      ↓
                              Playwright / Appium test files
                                      ↓
                              Pull request on your repo

ClaudeTest is a Model Context Protocol (MCP) server. Claude calls it as a tool during your conversation, using your GitHub token to read repos, trigger workflows, and open PRs.

---

Quickstart

Claude.ai (remote — easiest)

1. Open claude.ai → Settings → Integrations
2. Find ClaudeTest and click Connect
3. Authorize GitHub access when prompted

Claude Code (via npx)

npx @fasteningcode/claudetest --stdio

Add to .claude/settings.json:

{
  "mcpServers": {
    "claudetest": {
      "command": "npx",
      "args": ["@fasteningcode/claudetest", "--stdio"],
      "env": {
        "GITHUB_TOKEN": "ghp_your_token"
      }
    }
  }
}

Claude Code (from source)

git clone https://github.com/fasteningcode/claude-e2e-plugin
cd claude-enterprise-testing
pnpm install && pnpm build
export GITHUB_TOKEN=ghp_your_token
node packages/mcp-server/dist/index.js

Then point Claude Code at it in .claude/settings.json:

{
  "mcpServers": {
    "claudetest": {
      "command": "node",
      "args": ["packages/mcp-server/dist/index.js"],
      "env": {
        "GITHUB_TOKEN": "ghp_your_token"
      }
    }
  }
}

Self-hosted (Docker)

cp .env.example .env          # set GITHUB_TOKEN in .env
docker compose -f docker/docker-compose.yml up -d

Full guide: docs/self-hosted.md

---

Example conversations

Analyze a repo for coverage gaps

Analyze https://github.com/your-org/your-app for missing test coverage

Generate Playwright tests

Generate Playwright tests for the high-priority gaps you found

Generate Appium tests

Generate Appium tests for the untested screens in https://github.com/your-org/mobile-app

Diagnose a CI failure

Diagnose the failure in run 123456789 in https://github.com/your-org/your-app

Open a PR with generated tests

Open a PR with the generated tests on a branch called feat/claudetest

---

MCP Tools Reference

Tool	Input	What it does
analyze_repo	repo_url, branch?, paths?	Scans codebase, returns CoverageGap[] report
generate_tests	repo_url, file_paths[], test_type	Writes Playwright or Appium test files
run_tests	repo_url, workflow_id, branch?	Triggers GitHub Actions, polls until complete
diagnose_failure	repo_url, run_id	Parses CI logs, returns root cause + fix
create_pr	repo_url, branch, title, files[]	Creates branch, commits files, opens PR

---

Installation

npm i @fasteningcode/claudetest

npm: @fasteningcode/claudetest
MCP Registry: io.github.fasteningcode/claudetest

---

Repository structure

claude-enterprise-testing/
├── packages/
│   ├── mcp-server/              # MCP server entry point + all 5 tools
│   │   └── src/
│   │       ├── auth/            # GitHub token handling (per-request Bearer)
│   │       ├── github/          # Octokit wrapper
│   │       ├── tools/           # Tool handlers (one file per tool)
│   │       └── types.ts         # Shared Zod schemas + TypeScript types
│   ├── playwright-agent/        # Web test analysis + generation
│   │   └── src/
│   │       ├── analyze.ts       # Detects untested pages/components/routes
│   │       └── generate.ts      # Writes Playwright test files
│   └── appium-agent/            # Mobile test analysis + generation
│       └── src/
│           ├── analyze.ts       # Detects untested screens/navigation
│           └── generate.ts      # Writes Appium test files (iOS + Android)
├── fixtures/
│   ├── web-app/                 # Next.js fixture app (deliberately untested flows)
│   └── mobile-app/              # React Native fixture app (deliberately untested flows)
├── docker/
│   ├── Dockerfile               # Multi-stage production build
│   └── docker-compose.yml       # Self-hosted stack
├── docs/
│   ├── quickstart.md            # 5-minute setup guide
│   └── self-hosted.md           # VPC / Docker deployment guide
├── mcp.json                     # Claude.ai connector manifest
└── server.json                  # MCP Registry manifest

---

Tech stack

Layer	Technology
Plugin protocol	Model Context Protocol
Language	TypeScript (Node.js 20+)
Web testing	Playwright
Mobile testing	Appium + WebdriverIO
GitHub integration	Octokit
Schema validation	Zod
Test runner	Vitest
Package manager	pnpm workspaces

---

GitHub permissions required

Scope	Reason
repo read	Read source files and file tree
repo write	Create branches and commit test files
workflow write	Trigger and read GitHub Actions runs
pull_requests write	Open PRs with generated tests

---

Contributing

We welcome contributions! Please read CONTRIBUTING.md for the full guide.

Quick summary:

1. Fork the repo and create a branch: git checkout -b feat/your-feature
2. Make your changes with tests: pnpm test
3. Open a pull request against main

Great areas to contribute:

• New framework support (Vue, Angular, Flutter, Cypress)
• New CI integrations (GitLab CI, Bitbucket Pipelines, CircleCI)
• Cloud device farm support (BrowserStack, Sauce Labs, LambdaTest)
• Improved test generation heuristics

---

License

MIT