Self-maintaining E2E tests
Watch a team of agents inspect and fix broken tests, write bug reports, suggest and create new test scenarios.
Scaffolds a production-grade test project from scratch. Given a URL and a framework, the Architect sets up the folder structure, config, page objects, and install — then dispatches the Explorer and Writer to populate it with real tests.
- Creates the project skeleton with vendored templates per framework
- Installs dependencies and scaffolds page objects
- Orchestrates discovery and test generation
Learns your app by actually using it. The Explorer drives a real browser through your UI, reuses the flows your team has already captured, and discovers whatever it still needs — so tests are grounded in real user journeys, not guesses.
- Maps real user journeys — login, checkout, settings, anything your app does
- Builds on existing flow knowledge instead of duplicating work
- Feeds the Writer and Fixer a living map of your UI
Authors new tests in your framework's idiomatic style. Reads recorded flows and existing page objects, then produces tests that perform real interactions and assert on the resulting state — not just element visibility.
- Generates Playwright, Cypress, WebDriverIO, Selenium & SeleniumBase tests
- Prefers page objects over inlined selectors
- Writes state-changing assertions, not setup checks
Executes tests and reports pass, fail, and skip counts. Nothing else — no investigation, no code edits. Keeps the triage loop tight and honest: only the Runner decides if a test truly passes.
- Runs targeted tests, suites, or entire projects
- Surfaces skip diagnoses from XML reports automatically
- Provides the independent verification step after every fix
Diagnoses failures from pre-loaded DOM, network, and console evidence — then writes and verifies the fix in one step. Submits a diagnosis before any code change, so every fix is explainable and reviewable.
- Distinguishes app bugs, environment issues, and test-side flakiness
- Modes: code (fast fix), env (shell access), investigate (deep dive)
- Auto-reverts cross-file regressions on failure
The orchestrator. Coordinates the full triage: dispatches the Runner, assigns Fixers to failures with the right mode, escalates to the Explorer when flows are missing, and delivers the final verified summary.
- Plans the run, delegates, and verifies every fix with a fresh Runner pass
- Never edits code directly — only trusted agents touch files
- Saves new error patterns so your suite learns over time
Install E2Epilot
Copy and paste it in your terminal — a browser on localhost will open.
Requires Node 20+
Features
Work across multiple projects
Open several repositories in a single instance, or launch independent instances for fully isolated workspaces. Investigate parallel test suites without losing context.
Real bug detection
Goes beyond surface-level test failures. Distinguishes genuine application bugs from environment or test-side flakiness, and explains the failing path end to end.
Durable flaky-test fixes
Tracks intermittent failures across runs, identifies the underlying root cause, and applies long-term fixes — rather than masking symptoms with retries.
Learns your codebase
Recognises recurring failure patterns specific to your repository. Project-aware context sharpens diagnoses with every run, so frequent issues are resolved faster over time.
Reviewable fixes, never silent rewrites
Proposes targeted code changes your team reviews and merges. Every fix is explainable and reversible — no hidden edits, no changes you can't trace.
Enterprise-grade data privacy
Test execution and tool orchestration stay entirely on your machine. Only diagnostic prompts and error signals reach our AWS backend — your source code never leaves your environment, and nothing is ever used to train models.
Supported frameworks
Get in touch
Questions, feedback, or want to learn more? Reach out to us.
contact@e2epilot.com