• TRIAGE
  • ANALYSIS
  • RECOVERY
  • HUMAN

How to debug failed workflows with your AI

Connect your AI to Workload to find out why automations broke and fix them immediately.

Ask AI about this page

Short answer

How can I debug failed Workload workflows using AI?

Your AI pulls error logs and execution details from Workload to find the root cause of a failure. Once you understand the issue, you can tell your agent to retry the failed run. This gives you a direct path from error detection to recovery.

The outcomes.

Where every issue lands.

The AI handles the data retrieval and execution steps.

TRIAGE

Identify the error

The AI scans your recent logs to find exactly where a workflow stopped. You get a clear explanation of the failure instead of a raw error code.

ANALYSIS

Inspect the run

Your agent pulls the specific data from a failed instance. This provides the context needed to see if the issue was a bad input or a system timeout.

RECOVERY

Fix the execution

After you confirm the fix, the AI triggers a retry. This restarts the specific failed process in Workload so you don't have to do it manually.

HUMAN

Review the fix

You decide if the retry is safe or if the underlying logic needs a manual change. The AI waits for your signal before taking action.

The workflow

What your AI does when a workflow fails.

The AI moves from finding the error to attempting a fix.

  1. Find the failure

    The AI looks through your recent activity to find logs associated with failed runs.

    list_logs

  2. Get the details

    The AI pulls the specific data and state from a single failed instance to see what went wrong.

    get_execution

  3. Verify the fix

    The AI checks the logs again to ensure the error is no longer present or that the context is ready for a restart.

    list_logs

  4. Restart the run

    The AI sends a command to Workload to run the failed execution again.

    retry_execution

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 workflow names with your own.

Workload Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_Q3PtITVH3NUQK9iI8cHAkmiZAzWR2wFrFz6h5fAt/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 — Workload capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "workload-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_Q3PtITVH3NUQK9iI8cHAkmiZAzWR2wFrFz6h5fAt/mcp"
    }
  }
}
Copy into chat04

  • Show me the error logs for my last failed Workload execution.

  • Why did the execution with ID 98765 fail in Workload?

  • Check the logs for the 'Invoice Sync' workflow and tell me what happened.

  • Retry the failed execution from this morning in Workload.

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

Start here

Connect Workload once, then ask.

Link your account once and your credentials stay encrypted on our side. You can find the full walkthrough on the Connector page.

Connect Workload to your AI

FAQ

How this task behaves

  • 01

    Can the AI change my workflow logic?

    No. The AI can only read logs and retry existing executions. You must manually edit your workflow settings in Workload if the logic itself is broken.

  • 02

    Will the AI delete any of my data?

    No. This MCP does not have any tools for deleting logs, executions, or workflows.

  • 03

    Can I use this with Claude?

    Yes, as long as you are using an MCP-compatible client like Claude Desktop.

  • 04

    Does the AI retry every failure automatically?

    No. The AI will only retry an execution if you explicitly tell it to do so in the chat.

  • 05

    Can the AI create new workflows?

    No. The available tools are limited to listing logs, getting execution details, and retrying runs.

  • More questions about Workload? See everything the Workload Connector can do

Connect Workload to Claude, Cursor, ChatGPT & more