Back to portfolio
ProductIn development2025–Present

Hound

AI-native end-to-end test automation for authoring, running, and debugging browser tests with natural language and live browser control. Built end-to-end as a dashboard + Playwright execution engine + AI agent layer — so teams can ship durable tests without brittle selectors or slow manual QA loops.

Gallery

Record & author

Capture flows from a live browser session, refine steps with AI, and generate structured tests from natural language or production errors.

Execute & heal

Run tests with Playwright, AI locators/assertions, retries, recovery actions, and real-time streaming of every step.

Diagnose & observe

Pause-and-inspect debug, AI co-pilot, visual diffs, a11y/security/perf signals, and impact analysis when code changes land.

Product

Hound is an AI-native testing platform for teams shipping web apps. Instead of writing fragile selector-heavy scripts, you author tests in natural language, record from a live browser, or generate them from failures — then execute them with self-healing agents, live debug, and production-grade observability baked into every run.

Problem

Traditional E2E suites are slow to write, break on UI churn, and give weak failure signal. Teams bounce between brittle Playwright scripts, opaque CI logs, and separate tools for a11y, performance, and security — with little help turning production errors into coverage.

Solution

Hound unifies authoring, execution, and diagnosis: AI agents resolve elements and assertions, the engine retries and recovers, live view + co-pilot support mid-run debugging, and differentiators (impact analysis, chaos, coverage heatmaps, error → test) close the loop from production signal back to test coverage.

Highlights

  • Playwright execution engine with AI agents for locators, assertions, failure analysis, and recovery
  • Interactive recording from live browser view, plus AI step refinement after capture
  • Pause & inspect debug mode with AI co-pilot chat during runs
  • Visual regression (pixelmatch), HAR network capture, and video recording per run
  • Accessibility audits (axe-core / WCAG), Core Web Vitals budgets, and security scanning
  • Test impact analysis, chaos/resilience testing, and production-error → test generation
  • CLI runner, Playwright import/export, GitHub Actions integration, and plugin hooks (Slack, GitHub, Jira, Linear)
  • Branch-aware step cache, auth state save/load, scheduled monitoring, and S3/MinIO artifact storage

Specs

Type
AI product / developer tool
Role
Founder · Full-stack engineer
Company
Independent
Focus
AI-native E2E testing & live browser automation
Period
2025–Present
Stack core
Next.js · Playwright · Claude · Prisma
Status
Active development (Phase 10)
Surface
Dashboard · CLI · APIs · Live browser

How it works

  1. 1Author a test — Record from the live browser, write natural-language AI steps, import Playwright, or generate from a URL / production error.
  2. 2Run with agents — The engine executes via Playwright; AI agents resolve locators/assertions, cache hits speed repeats, retries + recovery handle flakiness.
  3. 3Watch live — Stream frames and step events in real time; pause to inspect, chat with the co-pilot, or step through failures.
  4. 4Close the loop — Review diffs, a11y/perf/security findings, impact analysis on code changes, and promote new coverage from real failures.

Achievements

  • Built a full-stack AI testing platform from foundation through production hardening (auth, Postgres, artifact storage, caching, retries)
  • Shipped 20+ step types spanning navigation, assertions, AI actions, auth, mocking, conditionals, a11y, and security
  • Delivered next-gen UX: live browser view, interactive recorder, debug toolbar, and AI co-pilot
  • Differentiated beyond “run Playwright in the cloud” with impact analysis, chaos testing, coverage heatmaps, and error → test generation
  • Added DX surface area: CLI (hound run / reporters), Playwright import/export, GitHub Actions, and a plugin architecture

Stack

TypeScriptNext.jsReactPlaywrightClaude APIPrismaPostgreSQLNextAuthTailwindshadcn/uiSSEAWS S3 / MinIOZodFramer Motion
Core StackFrontendBackend & DatabasesCloud & DevOpsData & Machine LearningDesign & DX