• LOG ANALYSIS
  • RUN STATUS
  • REPAIR PLAN
  • HUMAN REVIEW

How to debug autonomous agent logic with your AI

Connect Lindy to your AI client to pull real-time execution data. You'll see the exact thought process behind every decision.

Ask AI about this page

Short answer

How do I debug my AI agent's logic?

You can pull the raw reasoning logs and execution data directly into your chat. This lets you see the specific step where an agent's logic breaks down. You'll get a clear view of the following outcomes.

Where every issue lands.

Debug outcomes

Your AI uses Lindy's data to categorize the failure points.

LOG ANALYSIS

Identify logic gaps

The AI reads the execution history to find where the reasoning diverged from the goal. You get a clear explanation of the error.

RUN STATUS

Check execution state

The AI pulls the current status of a specific run. You'll know if a process is still running or if it hit a hard stop.

REPAIR PLAN

Fix the workflow

The AI suggests specific changes to your agent's instructions based on the logs. You get actionable steps to prevent the same error.

HUMAN REVIEW

Manual intervention

When the logs show an unrecoverable error, the AI flags it for you. You decide whether to restart the run or change the agent's permissions.

The workflow

What your AI does when a run fails.

Your AI acts as a debugger by pulling raw data from Lindy's execution engine.

  1. Fetch run details

    The AI pulls the specific metadata for a failed or stuck execution.

    get_run

  2. Extract reasoning logs

    The AI pulls the raw text of the agent's internal thought process and tool calls.

    get_run_logs

  3. Analyze the failure

    The AI compares the intended workflow against the actual logs to find the mismatch.

    get_run_logs

  4. Report findings

    The AI presents a summary of the error and the specific step that caused the hang.

    get_run

Try it

Copy these to start.

Use these prompts to begin debugging immediately.

Starting points

Swap the bracketed text for your actual Lindy run IDs or specific workflow names.

Lindy Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_c2F5Hk7TPFg1Cz9vjMKKvEPPvQH0HYkK2vL7pHhF/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 — Lindy capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "lindy-autonomous-ai-employees-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_c2F5Hk7TPFg1Cz9vjMKKvEPPvQH0HYkK2vL7pHhF/mcp"
    }
  }
}
Copy into chat04

  • Look at the logs for run ID [ID] and tell me why the agent stopped.

  • Check the status of run [ID] to see if it's still processing.

  • Review the reasoning logs for the [Workflow Name] agent to find the logic error.

  • Based on the logs from run [ID], what instruction should I change to fix this?

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

Start here

Connect Lindy once, then ask.

Just link your Lindy account to the MCP. Your credentials stay encrypted, and you can start debugging runs in any compatible client.

Connect Lindy to your AI

FAQ

Common questions

  • 01

    Can the AI fix the agent's code automatically?

    No. The AI analyzes the logs and suggests changes, but you must manually update the agent's logic in Lindy.

  • 02

    Can the AI restart a failed run?

    No. The AI can only read the status and logs. You have to trigger a new run yourself.

  • 03

    How does the AI see the agent's thoughts?

    It uses the get_run_logs tool to read the raw reasoning traces recorded during the execution.

  • 04

    Can I see the logs for every agent I've ever made?

    The AI can access logs for any run ID you provide through the get_run_logs tool.

  • 05

    Does the AI have permission to delete my workflows?

    No. This MCP only provides read access to run data and logs. It cannot modify or delete your Lindy configurations.

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

Connect Lindy to Claude, Cursor, ChatGPT & more