Beyond Copilots: New Amikoo agents use MCP to automate Playwright test repairs

#testing #softwaredevelopment #news #webdev

QA has officially become the industry bottleneck. Today, we are announcing a major evolution at MuukTest: the launch of Amikoo, an AI agent toolkit designed to give QA teams up to 8x more capacity to "fight fire with fire".

The Problem: The "Maintenance Tax" and Generic AI

Traditional no-code and low-code test automation is effectively dead in the age of LLMs. We’ve found that off-the-shelf AI agents are often too generic to test meaningfully because they lack context. This leads to the "Maintenance Tax"—the endless, boring cycle of manually fixing outdated tests every time the UI changes.

Our Solution: Context-Aware Agents

We built Amikoo (which means "friend" in Mayan) to be a trusted, knowledgeable partner for your QA function. Instead of a rigid platform, it’s a toolkit that learns and adapts to your specific software.

How it Works
Amikoo doesn't just guess what your software does. It leverages the Model Context Protocol (MCP) to pull precise, real-time data from your entire stack, including:

GitHub & CI/CDs: To understand code changes and architecture.
Jira & Pendo: To understand user context and high-priority flows.

The Toolkit
We are releasing specialized agents to handle the labor-intensive parts of the SDET workflow:

Amikoo No-Script Exploratory Agent: Maps all potential user flows to rapidly uncover coverage gaps and unintended actions.
Amikoo Test Repair Agent: This is the end of the maintenance tax. It monitors code changes and identifies broken tests, keeping a human in the loop only to verify if a change was intentional before repairing the code.

We believe you shouldn't be beholden to a single testing platform. Amikoo is designed to work harmoniously within existing Playwright, Claude, Cursor, and VS Code stack. Our goal for 2026 is "thoughtful testing"—letting humans stay in control of the strategy while AI handles the time-intensive tasks.