• FAILURE
  • CONTEXT
  • ACTION
  • HUMAN

How to debug failed test steps with your AI

Connect Zephyr Scale to your AI client to turn raw test failures into actionable insights. Once connected, your agent can pull execution details to pinpoint errors.

Ask AI about this page

Short answer

How can I use AI to find why my Zephyr Scale tests are failing?

Your AI uses the get_execution tool to pull the specific details of a failed test run. It then analyzes the step-by-step results to tell you exactly which condition wasn't met. You get a clear breakdown of the failure point.

The outcomes.

Where your debugging results land.

The AI processes the raw data and organizes the findings into these outcomes.

FAILURE

Pinpointed error steps

The AI identifies the specific step in the execution that triggered the failure. You get the exact reason the test stopped.

CONTEXT

Execution data summary

Your agent pulls the full context of the test run. This includes the parameters and environment details relevant to the failure.

ACTION

Suggested fixes

Based on the failure data, your AI suggests what might be wrong with the code or the test script itself.

HUMAN

Manual verification

You decide if the failure is a real bug or a flaky test. The AI provides the evidence, but you make the final call.

The workflow

What your AI does when a test fails.

The AI handles the heavy lifting of data retrieval and analysis.

  1. Fetch execution data

    Your agent reaches into Zephyr Scale to grab the specific details of the failed test run you're interested in.

    get_execution

  2. Isolate the failure

    The AI scans the retrieved steps to find the first instance where the actual result didn't match the expected result.

    get_execution

  3. Analyze step details

    It looks at the specific error messages and logs attached to that failed step to understand the root cause.

    get_execution

  4. Report findings

    The AI presents a concise summary of the failure, including the step name and the reason it failed.

    get_execution

Try it

Copy these to start.

Use these prompts to kick off the debugging process.

Starting points

Swap out the bracketed text with your actual Zephyr Scale execution IDs or test names.

Zephyr Scale (SmartBear) Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_Otgy8LaqzVGUEkWgebMQXIB54FzWeyFrkRQYbJXw/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 — Zephyr Scale (SmartBear) capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "zephyr-scale-smartbear-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_Otgy8LaqzVGUEkWgebMQXIB54FzWeyFrkRQYbJXw/mcp"
    }
  }
}
Copy into chat04

  • Get the details for Zephyr Scale execution ID 12345 and tell me why it failed.

  • Look at the last failed test in my Zephyr Scale project and find the broken step.

  • Summarize the failure reason for the test execution with ID 98765.

  • Check the execution details for the 'Login Flow' test and tell me if it's a data issue or a code issue.

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

Start here

Connect Zephyr Scale once, then ask.

Just link your account through the Vinkius dashboard. Your credentials stay encrypted, and you can immediately start asking your AI about your test results.

Connect Zephyr Scale to your AI

FAQ

Common questions

  • 01

    Can the AI fix the test for me?

    No. The AI can only read the execution data and suggest what might be wrong. You have to make the actual changes in your test suite.

  • 02

    Does the AI delete any test data in Zephyr Scale?

    No. This MCP only has permission to read execution data. It cannot create, update, or delete anything in your Zephyr Scale account.

  • 03

    Can I ask about multiple executions at once?

    Yes, you can ask your AI to compare the results of different execution IDs to find patterns in failures.

  • 04

    What if the test failed due to a network error?

    The AI will report the error message it finds in the execution data, which should include the network timeout or connection error.

  • 05

    Do I need to provide the execution ID every time?

    It helps. While you can ask for the 'last failed test', providing a specific ID makes the retrieval much faster and more accurate.

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

Connect Zephyr Scale (SmartBear) to Claude, Cursor, ChatGPT & more