Playwright tells you what failed. Now Currents tells you why, with AI Failure Analysis. On real CI failures, our specialized agent found the right root cause 81% of the time. An AI reading only the Playwright output found it 6% of the time. The agent combines failure information with execution steps, network logs, console logs and test behavior before and after a failure.

  • Get the root cause of a failed test without digging through logs
  • Get suggested fixes, each with a confidence level
  • Watch what the agent checks while it works: the steps, the requests, the screenshots
  • Copy the answer into a pull request, an issue or your coding agent's prompt
  • Available now on every plan
A Playwright error says what failed but not why. The agent goes through the test's steps and network requests and writes the likely cause with a suggested fix. On real CI failures it found the right cause 81% of the time, against 6% from the Playwright output alone.

When a test fails, expand Likely cause of the failure in the test's Insights panel. The agent reads the test steps, network logs, console logs and screenshots, and comes back with the type of failure, the root cause in a few sentences, and fixes to try.

The likely cause of a failed test

How we tested it

We took real CI failures with verified causes and gave each one to two AIs: one that reads only the Playwright output, and our agent.

Without Currents, the AI found the right root cause 6% of the time and a wrong one 52% of the time. With Currents, the agent found the right cause 81% of the time and a wrong one 3% of the time. Its first suggested fix addressed the cause in 76% of the failures, against 11% without Currents. It also suggested fewer fixes that only hide the bug, like a longer timeout or a retry: 6% against 17%.

Root cause and fix, with and without Currents

Enabling AI Failure Analysis

AI Failure Analysis sends test data to an AI provider, so an admin turns it on once for the whole organization: open the organization settings, select AI and data, and turn on AI assistance.

Want an agent to fix the failure in your code as well? Connect it to the Currents MCP server. It reads the same failure data and works in your repository.