• TRIAGE
  • ANALYSIS
  • REPAIR
  • HUMAN

How to debug failed automation flows with your AI

Connect Activepieces to your AI client to turn messy error logs into clear answers. Once connected, your agent can pull run data to find the root cause of any failure.

Ask AI about this page

Short answer

How can my AI help me find why an automation failed?

Your AI uses the Activepieces connector to pull specific execution logs and run history. It looks through the data to identify the exact step that errored out and explains why it happened. You get a clear diagnosis instead of digging through logs manually.

Where errors land.

The outcomes of a debug session.

The AI processes the execution data to give you these specific outcomes.

TRIAGE

Identify the broken step

The AI scans recent runs to find the specific component that failed. You'll know exactly which action caused the halt.

ANALYSIS

Explain the error

Your agent reads the error payload and translates technical jargon into plain English. You'll understand if it was a timeout or a bad payload.

REPAIR

Suggest a fix

Based on the failure data, your AI proposes changes to your logic or credentials. This gives you a starting point to get the flow running again.

HUMAN

Manual intervention

If the error is due to a missing API key or a changed schema, the AI tells you what you need to fix in the Activepieces UI. It won't change your settings for you.

The workflow

What your AI does when a flow fails.

The AI acts as a diagnostic layer between your error logs and your brain.

  1. Scan run history

    The AI looks through your recent automation attempts to find any that didn't complete successfully.

    list_flow_runs

  2. Inspect the failure

    Once a failed run is identified, the AI pulls the full execution details to see the exact error state.

    get_flow_run

  3. Isolate the trigger

    The agent checks if the failure happened at the start or during a specific action step.

    get_flow_run

  4. Report the cause

    The AI summarizes the data into a readable report so you can decide how to fix it.

    list_flow_runs

Try it

Copy these to start.

Use these prompts to kick off a debugging session.

Starting points

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

Activepieces Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_h2mPrUi2grDvdyHlWx6g8u5P9Bk6bXgJLfEMMCxF/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 — Activepieces capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "activepieces-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_h2mPrUi2grDvdyHlWx6g8u5P9Bk6bXgJLfEMMCxF/mcp"
    }
  }
}
Copy into chat04

  • Look at my recent Activepieces runs and tell me which one failed most recently.

  • Check the details for run ID 12345 and explain why the Google Sheets step failed.

  • Find all failed flows from the last hour and summarize the common error.

  • Analyze the execution data for my 'Customer Onboarding' flow and tell me if it's a data mapping issue.

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

Start here

Connect Activepieces once, then ask.

Just link your account once. Your credentials stay encrypted, and you can immediately start asking your AI about your automation logs.

Connect Activepieces to your AI

FAQ

How debugging works

  • 01

    Can the AI fix the broken flow automatically?

    No. The AI can only read the execution data and suggest fixes. You have to go into Activepieces to actually change the workflow.

  • 02

    Can the AI delete a failed run?

    No. The available tools only allow the AI to list and get run data. It cannot delete or modify any run history.

  • 03

    Does the AI see my API keys?

    No. The AI only sees the execution data and error messages returned by the flow runs.

  • 04

    Can I ask about a specific error message?

    Yes. Once the AI has pulled the data from a specific run, you can ask it to explain any part of that error.

  • 05

    How does the AI know which flows to look at?

    You tell it. You can ask it to look at all recent runs or a specific flow by name.

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

Connect Activepieces to Claude, Cursor, ChatGPT & more