AWT (AI Watch Tester)
Run AI-driven end-to-end UI tests with DevQA Loop and vision models that detect and suggest fixes for UI bugs.
Overview
AWT is a MCP server for the Ship phase that runs AI-powered E2E UI testing with DevQA Loop and Vision AI.
What is this MCP server?
- AWT (AI Watch Tester) MCP server — AI-powered E2E testing with DevQA Loop
- Vision AI detects UI bugs; workflow oriented toward auto-fix assistance
- PyPI package aat-devqa (published 1.6.2) with stdio transport, server version 1.6.3
- Provider modes: Anthropic (ANTHROPIC_API_KEY) or OpenAI (OPENAI_API_KEY) when selected
- Repository ksgisang/awt-skill on GitHub
- Server title AWT (AI Watch Tester), schema version 1.6.3
- PyPI identifier aat-devqa, listed package version 1.6.2
- 2 optional provider API key environment variables
Community signal: 1 GitHub stars.
What problem does it solve?
Manual UI regression is slow and flaky selector suites break whenever the design shifts.
Who is it for?
Indie builders with a web app or staging URL who want MCP-triggered E2E runs and AI-assisted bug triage before ship.
Skip if: Backend-only CLIs with no UI, or teams that forbid sending screenshots to cloud vision APIs.
What do I get? / Deliverables
Your agent drives vision-based E2E checks, surfaces UI defects, and loops through DevQA-style fix guidance faster than screenshot ping-pong.
- MCP-invoked E2E test runs with UI defect reports
- DevQA Loop iterations linking failures to fix suggestions
- Vision-model-backed validation before release
Recommended MCP Servers
Journey fit
How it compares
Vision E2E testing MCP server, not a static linter or unit-test runner skill.
Common Questions / FAQ
Who is AWT for?
Solo developers shipping web UIs who use Claude Code or Cursor and want MCP-controlled E2E tests with vision models.
When should I use AWT?
Use it in Ship/testing when you need pre-release UI regression and iterative fix hints via the DevQA Loop.
How do I add AWT to my agent?
Install aat-devqa from PyPI, configure stdio MCP in your agent, and set ANTHROPIC_API_KEY or OPENAI_API_KEY for your chosen provider.