The MCP AI Server That Gives Your Agents Eyes In The Browser

// Agent: discover a team and a GitHub-linked repo
const { teams } = await list_teams({ q: "acme" })
const { repos } = await list_repos({ q: "checkout-service" })

// Create the project
const project = await create_project({
  name: "Checkout E2E",
  platform: "web",
  teamUuid: teams[0].uuid,
  repoUuid: repos[0].uuid,
})

// Add a staging environment + admin credential
const env = await create_environment({
  projectUuid: project.uuid,
  name: "staging",
  url: "https://staging.checkout.acme.dev",
})
await create_credential({
  environmentId: env.uuid,
  label: "Admin",
  username: "admin@acme.dev",
  password: process.env.ADMIN_PW,  // goes in, never comes back out
  role: "admin",
})

// Run a browser test by role, not by password
await check_app_in_browser({
  url: "http://localhost:3000",      // localhost auto-tunneled
  description: "Add item to cart, apply coupon CART20, checkout, verify receipt page",
  environmentId: env.uuid,
  credentialRole: "admin",
})

// Later: inspect or cancel from history
const { executions } = await list_executions({ status: "running", limit: 5 })
if (executions[0].durationMs > 120_000) {
  await cancel_execution({ uuid: executions[0].uuid })
}