Tunnel AI - Natural Language to Frontend Tests
AI agent that turns plain-English instructions into automated Playwright tests using LangGraph, Browserbase, and Stagehand.
Technologies Used
About This Project
Tunnel AI transforms natural language into executable frontend tests. It generates, executes, and maintains tests end‑to‑end, eliminating the bottleneck of manual test creation while improving reliability for modern web apps.
The workflow is orchestrated with LangGraph as a multi‑agent system: a Planning agent structures requirements, a Generator produces Playwright code, an Executor runs tests at scale on Browserbase's cloud browsers, a Validator analyzes results, and a Self‑Healing agent fixes brittle selectors and timing issues.
Key features include self‑healing selectors, comprehensive test planning with edge cases, React‑aware strategies, parallel execution, and detailed reporting with screenshots and recordings. Tests can be run synchronously for fast feedback or enqueued for large suites.
Built with Python, LangGraph, Playwright, Browserbase, and Stagehand (browser automation SDK), Tunnel AI supports multiple LLM providers (OpenAI and Anthropic) and provides a simple CLI for developers.
Impact & Results
Currently in active development. Achieves 90%+ reduction in authoring time, executes 100+ tests in parallel on Browserbase, and auto-recovers broken selectors without manual edits.