Plugin + CLI (Claude Code)

The Plugin + CLI interface is the primary way to use Specwright. It installs as a project-local plugin and exposes 7 skills as slash commands inside Claude Code.

Setup

npx @specwright/plugin init

Then open your project in Claude Code.

Available skills

Typical workflow

1. Edit e2e-tests/instructions.js with your module config
2. Run /e2e-automate — pipeline runs until Phase 6
3. Review the plan and type yes to approve
4. Tests are generated and optionally healed
5. Commit the generated .feature + steps.js files

Running specific phases

# Explore only — useful when iterating on selectors
/e2e-plan http://localhost:3000/login

# Generate from an existing plan
/e2e-generate e2e-tests/plans/LoginPage-login-plan.md

# Heal a specific module
/e2e-heal @LoginPage

CI integration

Run tests in CI without Claude Code:

npx bddgen && npx playwright test --project main-e2e

Specwright generates standard Playwright BDD tests — CI only needs Playwright, not Claude. The AI pipeline writes and maintains tests; it doesn't run them in production.

Project structure after install

your-project/
├── .claude/
│   ├── agents/          ← 8 AI agent definitions
│   ├── skills/          ← 7 pipeline skills
│   └── agent-memory/    ← persisted selectors + fix patterns
├── .mcp.json            ← Playwright MCP server config
└── e2e-tests/
    ├── .env.testing     ← credentials (gitignored)
    ├── features/playwright-bdd/
    │   ├── shared/      ← globally-scoped steps
    │   ├── @Modules/    ← generated module tests
    │   └── @Workflows/  ← generated workflow tests
    ├── playwright/
    │   ├── fixtures.js  ← custom BDD fixtures
    │   └── generated/seed.spec.js
    └── utils/
        ├── stepHelpers.js
        └── testDataGenerator.js