E2E Tests MCP Guide

QE E2E Tests V2 — standard Playwright UI/API generation, codegen, k6 performance, and failure healing. Not the @qe/blocks server.

Overview

The E2E Tests V2 MCP generates Playwright E2E tests that follow INDG conventions: AAA, TDD, POM or Screenplay where appropriate, semantic selectors, robust waits, and non-destructive generation. Sibling tools cover API tests, codegen, k6 performance, spec-driven generation, and healing failed Playwright tests.

This server uses standard @playwright/test. For Bloomberg Law @qe/blocks tests, use QE Blocks Tests V2 instead. Prefer adding .qe/ to .gitignore when healing or gap-style reports are written under .qe/reports/.

URLs and JSON

Mount: /mcp/v2/e2e_tests. Config key: qe_e2e_tests_v2. Legacy V1 mounts under /mcp/qe_e2e_tests are deprecated — use V2 only.

Environment	URL	Transport	Status
Development (V2)	`https://qe-mcp.bindg.com/mcp/v2/e2e_tests`	http	V2
Production (V2)	`https://qe-mcp.bindg.com/mcp/v2/e2e_tests`	http	Production

Cursor — `~/.cursor/mcp.json`

{
  "mcpServers": {
    "qe_e2e_tests_v2": {
      "url": "https://qe-mcp.bindg.com/mcp/v2/e2e_tests",
      "transport": "http"
    }
  }
}

VS Code — `.vscode/mcp.json`

{
  "servers": {
    "qe_e2e_tests_v2": {
      "url": "https://qe-mcp.bindg.com/mcp/v2/e2e_tests",
      "type": "http"
    }
  },
  "inputs": []
}

Which tool for which job

You want…	Tool
New UI E2E, repo scan, standards, folder layout	`e2e_core` (start here)
Record the journey, then clean up	`e2e_codegen` (with `e2e_core`)
HTTP / API / contract tests	`e2e_api` (with `e2e_core`)
Turn a written spec into tests	`e2e_spec_keyword` (with `e2e_core`)
Load / performance (k6, not browser click-load)	`e2e_performance`
Failed tests — why they fail and safe fixes	`playwright_test_analyzer`

Add to Cursor

Open Cursor Settings.
Search for Tools & MCP.
Choose New MCP Server.
Paste the Cursor JSON above, or add a single entry.
Save and confirm the server appears in Installed MCP Servers.
Start the server and run a smoke-test prompt.

Add to VS Code

Open the Command Palette (Cmd+Shift+P or Ctrl+Shift+P).
Search for MCP: Add Server and select HTTP.
Enter the URL https://qe-mcp.bindg.com/mcp/v2/e2e_tests.
Provide an alias such as qe_e2e_tests_v2.
Save globally or to the workspace and start the server.

Playwright E2E workflow

The MCP generally follows this workflow when generating E2E tests:

1. Repository analysis

Identify framework, stack, routing, state management, and build tooling.
Find existing Playwright config.
Locate current test structure, page objects, fixtures, and utilities.
Identify testable user flows and core features.

2. Test architecture proposal

Reuse existing conventions when tests already exist.
Propose a feature-based layout when no tests exist.
Include POM or Screenplay patterns when appropriate.
Define fixture, parallelization, and retry strategy.

3. Test generation

Create new test files only.
Do not overwrite config, page objects, fixtures, utilities, or existing tests.
Use semantic selectors.
Follow Arrange / Act / Assert.
Prefer stable waits and auto-waiting assertions.
Propose data-testid additions when selectors are missing.

Rules and best practices

When generating Playwright tests, the assistant behaves like a senior Playwright architect and follows:

TDD and AAA consistently.
Maintainable architecture using POM or Screenplay where appropriate.
Repo-first workflow before writing tests.
Non-destructive generation — no overwrites of config, page objects, fixtures, utilities, or tests.
Reliable locator strategy: prefer data-testid, then roles, labels, text, placeholders.
No brittle selectors such as nth-child.
No arbitrary waits such as waitForTimeout.
Robust Playwright waits and assertions (waitForURL, waitForResponse, auto-waiting matchers).
API mocking where appropriate (page.route(), request fixture).
Typed, production-grade code, especially in TypeScript projects.

Performance (k6) and API testing

Use e2e_api (with e2e_core) for HTTP/API tests and e2e_performance for k6 load scripts — not browser click-load. Keep UI, API, and load goals separate when they differ.

API: Using MCP server qe_e2e_tests_v2, generate API tests from our OpenAPI for the billing endpoints (Playwright request). Don’t invent endpoints.

k6 scenario: Using MCP server qe_e2e_tests_v2, create a k6 smoke and load script for the search API with our SLO thresholds.

Performance platform (EC2)

e2e_performance generates k6 scripts in your repo. To run those scripts at team scale on shared AWS infrastructure — Terraform-managed EC2 runners, SSM deploy, and S3 report upload — use the Performance Tool Guide.

Performance Tool Guide — key sections

MCP workflow for k6 generation — from prompt to k6-tests/ layout
Step-by-step EC2 execution — Terraform, run-tests.sh, and S3 verification
Platform overview — what the shared k6 EC2 framework provides
Prerequisites — AWS CLI, Terraform, k6, and repo access

End-to-end path: generate k6 with e2e_performance, validate locally, then promote to the shared EC2 platform for scaled runs and S3 report review.

You need…	Where to go
Generate a k6 script via MCP	Performance guide — MCP workflow
EC2 run and S3 upload	Performance guide — step-by-step
Repo, Terraform, and AWS setup	Performance guide — prerequisites

Example prompts

UI / E2E: Generate Playwright E2E tests for the login and save-search flows. Follow our existing tests/e2e layout. Using MCP server qe_e2e_tests_v2.

Repo discovery: Using MCP server qe_e2e_tests_v2, analyze this repository's existing Playwright setup. Output a concise inventory of config, directories, fixtures, page objects, utilities, and what should be reused.

Generate new tests: Using MCP server qe_e2e_tests_v2, generate NEW Playwright E2E tests for the login feature using AAA and POM. Do not overwrite any existing files. Show the exact file paths you will create.

From a spec: Here’s the feature spec for export filtered records — generate Playwright tests from it; ask before adding keyword boilerplate.

Codegen integration

Use e2e_codegen with e2e_core for record-first workflows.

Example: transform codegen output

Help me record the profile-update flow with Playwright codegen, then turn it into clean POM tests. Using MCP server qe_e2e_tests_v2.

Example: record a specific flow

Set up codegen for our base URL and walk me through recording search.

The assistant should:

Ask once before launching the headed recorder.
Guide you through recording the journey.
Transform raw codegen output into structured test files (AAA, semantic selectors, POM).
Save tests to a feature-specific test directory.
Replace brittle CSS selectors with semantic selectors where possible.

        Tip: After generation, run the suite — generation alone
        does not prove the tests pass. Name the feature, environment, and existing test folder.
      

UI Mode alternative

npm run test:ui

Failure healing

Use playwright_test_analyzer after Playwright failures — not for writing new tests from scratch. It classifies root cause, optionally inspects the live UI (Playwright browser MCP), writes .qe/reports/e2e-test-healing-{date}.md, and asks for confirm-before-apply fixes.

Heal CI failures: These Playwright tests failed in CI — investigate and propose safe fixes (confirm before apply). Using MCP server qe_e2e_tests_v2.

Cross-platform npm scripts

The rules ensure generated npm scripts work on Windows, macOS, and Linux:

Use cross-env instead of export for environment variables.
Use shx instead of rm, cp, or ln.
Use --prefix instead of cd && patterns.
Use forward slashes in paths (Node.js normalizes automatically).
Use npm-run-all for sequential / parallel script orchestration.

Validate the connection

Smoke test: Use MCP server qe_e2e_tests_v2 to list available tools and capabilities and provide three example prompts for Playwright E2E workflows.

If the server fails to respond, see BBVPN proxy configuration on the hub.

← Back to MCP Server Guide ← Back to Quality Engineering Tools Suite