• LOGS
  • STATUS
  • ACTION
  • HUMAN

How to debug failed workflows with your AI

Connect Conductor to your agent to handle troubleshooting through chat. You stop digging through dashboards and start fixing issues directly.

Ask AI about this page

Short answer

How do I debug failed workflows with my AI?

Your agent uses Conductor tools to fetch specific workflow data and task logs to find errors. Once the cause is clear, you can tell your agent to retry the workflow immediately. This breakdown shows how the agent moves from discovery to resolution.

Where every issue lands.

Troubleshooting outcomes

The agent categorizes every Conductor event into one of these four states.

LOGS

Error identification

The agent reads the raw output from failed tasks. You get a plain English explanation of why the process stopped.

STATUS

Workflow state

The agent checks the current progress of a specific execution. You see exactly which step is stuck or broken.

ACTION

Automated retries

The agent triggers a new attempt for a failed workflow. You fix the execution without touching the Conductor console.

HUMAN

Manual intervention

The agent flags issues that require code changes or infrastructure fixes. You decide when to step in and when to let the agent retry.

The workflow

What your AI does when a workflow fails.

The agent acts as your Conductor operator, pulling data and executing commands based on your questions.

  1. Locate the failure

    The agent finds the specific execution you are asking about by pulling its current state.

    get_workflow

  2. Inspect task details

    It looks at the individual steps within that workflow to see which one hit an error.

    get_workflow_tasks

  3. Read error logs

    The agent pulls the specific stack trace or error message from the failed task to understand the root cause.

    get_task_logs

  4. Execute a retry

    After you confirm the fix, the agent restarts the workflow to clear the error.

    retry_workflow

Try it

Copy these to start.

Use these prompts to begin troubleshooting immediately.

Starting points

Swap out the workflow IDs and task names with your actual Conductor identifiers.

Conductor (Netflix OSS) Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_TUjFjxN9H3g5Ba44H63GIIO4Av31lZmg9U3GcY2P/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 — Conductor (Netflix OSS) capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "conductor-netflix-oss-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_TUjFjxN9H3g5Ba44H63GIIO4Av31lZmg9U3GcY2P/mcp"
    }
  }
}
Copy into chat04

  • Find the status of workflow 'order-process-123' and tell me why it failed.

  • Get the logs for the last failed task in workflow 'payment-sync-99'.

  • List all tasks for workflow 'data-pipeline-abc' to see where it is stuck.

  • Retry the workflow 'user-onboarding-456' now that I've fixed the service.

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

Start here

Connect Conductor once, then ask.

Just link your Conductor instance to Vinkius. Your credentials stay encrypted, and you can start debugging through your favorite AI client right away.

Connect Conductor to your AI

FAQ

How the Conductor debugger behaves

  • 01

    Can the AI delete a workflow?

    No. There is no tool provided to delete workflows or tasks.

  • 02

    Will the AI change my workflow definition?

    No. The agent can only read data and trigger retries; it cannot modify the underlying Conductor logic or code.

  • 03

    Can I use this to create new workflows?

    No. This specific set of tools is for inspecting and retrying existing workflows.

  • 04

    Does the AI need the Conductor UI open?

    No. Once connected, the agent performs all lookups and retries through the MCP.

  • 05

    Can the agent fix code errors automatically?

    No. The agent can identify the error in the logs, but you must fix the actual code or environment before telling the agent to retry.

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

Connect Conductor (Netflix OSS) to Claude, Cursor, ChatGPT & more