Claude MCP · Playwright-grade E2E · early access

Record your E2E tests in one chat.
Replay them 100× cheaper.

Claude MCP test automation for Playwright-grade E2E. Use Claude where it shines — designing the test once. Replay from saved snapshots: far fewer LLM tokens, faster runs, full history and video reports. Self-healing on broken Playwright and Cypress selectors.

Connect Claude — Free Book a demo
  • 2 minto first test
  • 100×fewer tokens / replay
  • 100%runs logged + video
  • 10×faster than vision
Speaks every modern AI client Claude Desktop Claude.ai (web) Cursor Continue.dev Cline Any MCP-compatible client
Why this matters

Three things change
the moment you connect.

Cut the LLM bill on E2E

Pay Claude once
to design. Replay 100× cheaper.

Claude authors the test. Playwright replays from saved snapshots — fewer tokens, faster runs, full reports on every replay.

Marginal cost per replay · 100 steps

Claude Computer Use
~$3.20
Cypress / Playwright
~$0.12
E2Easy in our cloud
~$0.02
E2Easy on your machine
$0.00

Computer Use cost is LLM tokens on every step. The rest is CI compute. After authoring, replays only call Claude when self-healing fires.

Real selectors, not screenshots

Speak DOM, not pixels.

Stable selectors, ARIA, xpath, locators — pulled from the live DOM. Never invented from memory. 

// inspect_page → element
{
  "selector": "button[data-test='login-submit']",
  "aria":     "Log in",
  "xpath":    "//main/form/button[2]",
  "locator":  "role=button[name='Log in']",
}
Lives in a real suite

Auditable. Versioned. CI-ready.

Same dashboard as your manual recordings. Video, logs, scheduling, CI hooks included.

Sign-up flow Passing
Checkout — guest Passing
Password reset Flaky 82%
Admin invite Passing
How it works

From a chat message
to a regression test.

Two minutes to set up. Then your suite grows from chat instead of from sprints.

Connect via OAuth

One-click MCP connector for Claude Desktop, Claude.ai, Cursor. No static keys.

Revocable any time

Describe the flow

"Cover sign-up, add to cart, checkout as guest." Claude confirms before saving.

Plain English, no DSL

Claude inspects the page

Real Playwright browser. Stable selectors, ARIA, xpath — pulled from live DOM.

Real DOM, real assertions

Replay anywhere

Local, your CI, or our cloud. Deterministic. Video, logs, full run report.

100× fewer tokens / replay
Real prompts, real outcomes

What you can ask Claude
once you connect.

Six prompts pulled from real teams. Each one is a single message — Claude does the rest.

Author from scratch

Cover the entire onboarding wizard for a new tenant on app.acme.com.

17 steps, 4 assertions. One test, saved.

LLM · authoring
$0.07
LLM · per replay
$0.00
Self-heal a regression

My checkout test failed last night. Look at it and fix the broken step.

Patches only the failing step. The other 11 stay untouched.

LLM · this fix
$0.01
Engineer time
0 min
Bulk refactor

Move every test that hits /admin to the new /workspace path.

14 tests rewritten in one chat. Claude confirms each change.

LLM · 14 edits
$0.18
LLM · per replay
$0.00
Triage flaky tests

Which tests have been flaky this week? Suggest the next two to fix.

Ranked by pass-rate, with the runs that broke them.

LLM · this triage
$0.02
Saved engineer hours
~3 hr
Run from chat

Run all checkout tests against staging and tell me which ones regressed.

Parallel dispatch. Pass / fail / slower — with run links.

LLM · orchestration
$0.02
LLM · per replay
$0.00
Onboard juniors / PMs

Add coverage for the new pricing page. Don't break anything else.

PMs author tests. Claude confirms before saving. No code.

LLM · authoring
$0.04
Coding required
None
// the math

Tune the sliders.
See your real bill.

Drop in your numbers — tests, replays per test per year, engineer hourly rate. The calculator updates live.

Same intelligence in all three rows. Only one keeps charging on every replay.

Approach LLM / yr Eng. time / yr Total / yr
Claude Computer Use $0 0 hr $0
Cypress / Playwright $0 0 hr $0
E2Easy + Claude $0 ~0 hr $0

You save $0 a year vs Computer Use, $0 vs hand-written Cypress.

Connect Claude — Free Talk to founder

Computer Use cost = LLM tokens at ~$3.20 per 100-step replay. E2Easy LLM cost = $0.04 per authoring + occasional self-heal (~$0.01). CI compute on top: $0 local · ~$0.02/run in our cloud.

The toolset Claude gets

A focused MCP API.
No prompt-engineered hacks.

10 deliberate verbs. Claude composes them into anything from one-liner to full suite.

inspect_page(url) → Element[]// real browser, interactive elements
find_element(description, url) → Element// plain-English locator
create_test(name, steps[]) → Test// save a new E2E test
list_tests(filter?) → Test[]// search workspace
get_test_steps(test_id) → Step[]// read step list
replace_test_steps(test_id, steps[]) → Test// patch one step or rewrite
update_test(test_id, patch) → Test// rename, move, tag
delete_test(test_id) → void// destructive — confirms first
run_test(test_id, env?) → RunId// dispatch to runner
list_test_runs(test_id?) → Run[]// run history
OAuth 2.1 + PKCE. Short-lived tokens. Revocable.
Destructive verbs require Claude to confirm first.
Side by side

The same intelligence,
without the runtime tax.

Computer Use reasons. We make reasoning durable, deterministic, replayable.

Claude Computer Use

Agent drives the browser

  • Brilliant at reasoning
  • Pays LLM tokens every run
  • Probabilistic, slow per step
Cypress / Playwright

Engineer writes the code

  • Deterministic, fast
  • Hours of code per flow
  • Manual fixes on UI changes
This is the one E2Easy + Claude

Author once, replay anywhere

  • One chat — saved as a recorded test
  • Replay locally, on your server, or in our cloud
  • 100× fewer tokens per replay
  • Full history, video and run reports
Case study · Forge.io

From 4 hours per sprint
to 22 minutes.

How a 12-engineer team at Forge.io ran their entire E2E maintenance off Claude MCP — and cut LLM spend 87× in the process.

E2E suite size
47 tests · ~2,800 steps
Replays / month
~9,400
Stack before
Cypress + manual fixes
Stack now
Claude Desktop + E2Easy MCP + Playwright

The problem

Every UI refactor broke 3-7 selectors. Two engineers on rotation spent ~4 hours per sprint hunting and patching them. Bug-fixing PRs queued behind selector PRs. They'd evaluated Claude Computer Use but the per-replay LLM bill was a non-starter for nightly + per-PR runs.

The flip

They switched in one afternoon: connected Claude Desktop to E2Easy via MCP, re-recorded the broken tests in chat, then pushed all 47 to the platform. The team kept Cypress for the unit-level UI tests and moved the cross-page flows to E2Easy.

The numbers · 90 days in

  • 22 min selector maintenance per sprint (was 4 hr)
  • $3.10 total LLM spend / month (Computer Use estimate: $270)
  • 11 → 0 selector-fix PRs per sprint blocking releases
  • 100% runs with video + step logs available in dashboard
A selector breaks, Claude patches that one step. The rest of the suite keeps running. Self-healing without re-prompting the whole flow — and the run history actually means something now, because every replay is the same path.
Rishi Tandon · Staff Engineer, Forge.io
Try the same setup — Free Talk to founder about your suite →
Pricing

Bring your own Claude.
The connector is on us.

Full MCP access. Unlimited replays. No credit card. Founder pricing locked in.

Early access
$0 / month, all features
  • Claude MCP connector
  • Unlimited replays
  • Up to 100 saved tests
  • 1,000 test runs / month
  • Video on every run
  • OAuth-secured access
  • Run history & logs
  • Priority support
Connect Claude — Free

No credit card · Cancel anytime · Founder discount when paid plans launch

Guides & tutorials

Claude MCP test automation,
step by step.

From Playwright MCP setup to self-healing Cypress selectors with Claude.

Setup · 5 min

Connect Claude to Playwright via MCP

OAuth setup for Claude Desktop, Claude.ai, Cursor. First test in 5 min.

Stack
Claude · MCP · Playwright
Difficulty
Beginner
Self-healing · 8 min

Auto-fix broken Cypress selectors with Claude

UI refactor breaks one step. Claude patches that one step. Suite stays green.

Stack
Claude · Cypress · Playwright
Outcome
Suite green again
E2E authoring · 10 min

Self-healing E2E tests with Claude MCP

Author once. Self-heal broken selectors on demand. 100× fewer tokens per replay.

Stack
Claude MCP · Playwright
Saving
~80% maintenance
CI · 6 min

Run Playwright tests in CI with snapshot replay

GitHub Actions, GitLab CI, your server or our cloud. Logged, with video reports.

Stack
Playwright · CI
LLM cost
$0 / replay
Concept · 4 min

MCP test automation explained

Why Claude e2e testing via MCP beats Computer Use — and where it fits in your stack.

Format
Reading
For
QA leads · CTOs
Generation · 7 min

Automated test case generation with Claude Code

From Jira ticket, Figma flow or bug report → Playwright-grade test in your suite.

Stack
Claude Code · Playwright
Coding
None
Frequently asked

Everything else
you might be wondering.

What is MCP and why does E2Easy use it?
Anthropic's Model Context Protocol — the open standard for plugging LLMs into real tools. E2Easy's MCP server gives Claude native verbs for recording, inspecting and replaying browser tests.
Does Claude run my tests every time?
No. Claude authors once. Playwright replays from the saved snapshot — locally, in CI, or in our cloud. Self-healing fires only on the broken step.
Is this the same as Claude Computer Use?
No. Computer Use calls the LLM on every step. E2Easy uses Claude to author the test once from the live DOM, then replays deterministically from snapshots — far fewer tokens per replay, with full video reports.
Do I need an Anthropic API key?
No. Bring your own Claude account (Desktop, Claude.ai, Cursor — anything MCP-compatible). OAuth-based, no static keys.
Will my tests still work without Claude?
Yes. Tests live on Playwright. Claude is the author, not a runtime dependency.
Can other MCP-compatible clients use this?
Yes. The E2Easy MCP server speaks standard MCP over OAuth. Claude Desktop, Cursor, Continue.dev, Cline and similar clients all connect with the same endpoint.
How does Claude know which selectors to use?
It doesn't guess. inspect_page and find_element visit the page in a real Playwright browser and return stable selectors, ARIA, xpath, locators — straight from the DOM.
How is this different from Microsoft Playwright MCP?
playwright-mcp is a raw browser bridge — every replay calls the LLM on each step. E2Easy stores tests as Playwright-grade recordings: replay is deterministic with run history and video reports, and self-healing fires only on the failing step.
Can I use this to fix broken Cypress selectors with Claude?
Yes. CSS, ARIA, xpath, role-based locators — all supported. After a UI refactor Claude reads the live DOM via MCP and produces a drop-in selector fix. Only the failing step is patched.
Does this count as AI test automation or self-healing test automation?
Both. AI test automation for authoring, self-healing test automation for maintenance. Claude patches only the failing step.
What about security?
OAuth 2.1 + PKCE. Short-lived tokens. Revocable any time. Test data stays in your workspace.

Use Claude to design
your tests. Not to run them.

Two minutes to connect via MCP.
Author once. Replay 100× cheaper, fully logged, with video reports.

Connect Claude — Free Book a demo

No credit card · Cancel anytime · OAuth 2.1 + PKCE