• LOGS
  • STATE
  • HEAL
  • HUMAN

How to debug failed GUI tests with your AI

Connect ContextQA to your AI client to turn raw test logs into actionable debugging insights. Once connected, your agent handles the heavy lifting of scanning failures.

Ask AI about this page

Short answer

How do I debug failed GUI tests with my AI?

Your AI pulls execution data and specific test cases to pinpoint exactly where a GUI interaction broke. It then analyzes these failure points to suggest where the test suite needs healing. You get a clear breakdown of the error and the state of the application at the time of failure.

Where every issue lands.

Debugging outcomes

The AI categorizes every failure it finds into one of these four buckets.

LOGS

Failure analysis

The AI uses execution data to explain why a specific step failed. You get a plain-English reason instead of a stack trace.

STATE

UI state snapshots

The agent looks at the specific test case to see what the UI looked like during the crash. This helps you see if an element was missing or obscured.

HEAL

Healing suggestions

The AI identifies where a selector changed or a timing issue occurred. It points out exactly where the test needs a fix to pass again.

HUMAN

Manual review required

When the failure is a legitimate bug in your app rather than a flaky test, the AI flags it for you to investigate manually.

The workflow

What your AI does when the failure arrives.

The AI moves from broad execution lists to specific failure details without you needing to hunt through logs.

  1. Identify failures

    The AI looks through your recent test runs to find any executions that didn't pass.

    list_executions

  2. Pull execution data

    Once a failure is found, the agent grabs the specific details of that run to see the error logs.

    get_execution

  3. Inspect the test case

    The AI pulls the underlying test case data to compare the expected behavior against what actually happened.

    get_case

  4. Summarize the gap

    The agent compares the failed execution against the test case to explain the discrepancy.

    get_executionget_case

Try it

Copy these to start.

Use these prompts to kick off the debugging process.

Starting points

These are opening lines. Swap out the specific execution IDs or case names with your actual data.

ContextQA Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_arIg6fOatFWYf4ywr6Asm6h013dTDEa83NCzuGoP/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — ContextQA capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "contextqa-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_arIg6fOatFWYf4ywr6Asm6h013dTDEa83NCzuGoP/mcp"
    }
  }
}
Copy into chat04

  • List my recent test executions and tell me which ones failed.

  • Look at execution ID 12345 and explain why the GUI test failed.

  • Check the test case for 'Login Flow' and compare it to the last failed execution.

  • Analyze the failure in execution 9876 and tell me if I need to heal the test selectors.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect ContextQA once, then ask.

Just follow the link to connect your account. Your credentials stay encrypted, and you can immediately start asking your AI about your test results.

Connect ContextQA to your AI

FAQ

How this task behaves

  • 01

    Can the AI fix my code automatically?

    No. The AI identifies the failure points and suggests where healing is needed, but you must apply the actual code changes yourself.

  • 02

    Can the AI delete my test results?

    No. The tools only allow the AI to read execution and case data. It cannot modify or delete any data in ContextQA.

  • 03

    Does the AI need my login password?

    No. Vinkius manages the connection securely so your AI client can access the data without you sharing passwords in chat.

  • 04

    Can I ask about a specific test case?

    Yes. You can use the tools to pull details for any specific test case to investigate its logic or history.

  • 05

    What happens if a test fails due to a real app bug?

    The AI will report the failure based on the execution data, but it will flag the issue as a potential application bug rather than a test flake.

  • More questions about ContextQA? The Connector page answers them. See everything the ContextQA Connector can do

Connect ContextQA to Claude, Cursor, ChatGPT & more