• TRACE
  • ERROR
  • LOGIC
  • HUMAN

How to debug agentic execution traces with your AI

Connect LangSmith to your agent to pull real-time telemetry. You can then inspect exactly why a tool call failed or where a reasoning chain went off the rails.

Ask AI about this page

Short answer

How can I debug my AI's reasoning and tool calls?

Your AI pulls specific run data from LangSmith to show you the exact sequence of events in a trace. You'll see the inputs, outputs, and nested logic for every step in the execution path.

Debugging outcomes

Where your telemetry lands.

The AI organizes LangSmith data into clear debugging paths.

TRACE

Full execution paths

The AI pulls the complete sequence of events. You see every nested call and decision point in order.

ERROR

Failure points identified

The AI isolates specific runs that failed. It points you to the exact tool call or prompt that caused the error.

LOGIC

Reasoning gaps found

The AI compares expected outcomes against actual traces. You get a clear view of where the agent's logic diverged.

HUMAN

Manual review needed

The AI flags complex or ambiguous traces. You step in to inspect the raw data when the logic is too messy for automated summary.

The workflow

What your AI does when <the thing arrives>.

The AI acts as a bridge between your chat interface and your LangSmith telemetry.

  1. Locate the project

    The AI finds the correct LangSmith project to start the investigation.

    list_projects

  2. Find the specific run

    The AI scans the project to find the exact execution trace you're looking for.

    list_runs

  3. Pull trace details

    The AI fetches the full telemetry, including nested tool calls and reasoning steps.

    get_run

  4. Analyze the path

    The AI reviews the retrieved data to explain the agent's behavior.

    get_run

Try it

Copy these to start.

Use these prompts to begin your debugging session.

Starting points

Replace the project or run names with your actual LangSmith identifiers.

LangSmith Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_ixIJrwVWDWlC1bEOf4IyffAPgAEw302Ns6BE1oLo/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 — LangSmith capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "langsmith-llm-observability-hub-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_ixIJrwVWDWlC1bEOf4IyffAPgAEw302Ns6BE1oLo/mcp"
    }
  }
}
Copy into chat04

  • List all my LangSmith projects so I can pick one to debug.

  • Find the most recent failed runs in my 'production-agent' project.

  • Get the details for run ID 'abc-123' and explain why the tool call failed.

  • Look at the last five runs in my 'test-suite' project and tell me if the reasoning path is consistent.

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

Start here

Connect LangSmith once, then ask.

Just link your LangSmith account. Your credentials stay encrypted, and you can start asking questions about your traces immediately.

Connect LangSmith to your AI

FAQ

How this task behaves.

  • 01

    Can the AI change my LangSmith project settings?

    No. The AI can only read telemetry and project lists. It cannot modify your LangSmith configurations or delete projects.

  • 02

    Can the AI fix the code that caused the error?

    No. The AI identifies where the error happened in the trace, but you must manually fix the code in your own development environment.

  • 03

    Does the AI see all my data?

    The AI only sees the specific run data you ask it to retrieve using the provided tools.

  • 04

    Can I ask about runs from a different project?

    Yes, as long as you first use the tool to list and identify that project.

  • 05

    Will the AI delete old traces to save space?

    No. The AI has no capability to delete or manage the storage of your LangSmith data.

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

Connect LangSmith to Claude, Cursor, ChatGPT & more