- TRACE
- LATENCY
- ERROR
- HUMAN
Short answer
How can I debug my agent's tool calls?
Your agent uses Metorial to pull granular execution traces that pinpoint where tool calls fail or hang. You'll get a clear view of the inputs and outputs for every step.
Debugging outcomes.
Where your data lands.
The AI sorts trace information into specific categories for you.
TRACE
Identify logic gaps
The AI pulls trace logs to show where the agent's reasoning diverged from the intended path.
LATENCY
Find slow calls
The AI checks execution timing to highlight which specific tools are causing delays.
ERROR
Pinpoint failures
The AI isolates the exact tool call that returned an error and explains why it happened.
HUMAN
Manual review needed
The AI flags traces that require you to step in and manually correct the agent's logic or parameters.
The workflow
What your AI does when a debugging request arrives.
The AI acts as the bridge between your agent's logs and your chat window.
Scan recent activity
The AI looks through recent execution logs to find the relevant tool interactions.
metorial_list_tracesIsolate specific traces
The AI selects the exact trace that matches your error or latency report.
metorial_get_trace_detailsInspect tool data
The AI reads the specific inputs and outputs used during the failed or slow call.
metorial_get_trace_detailsSummarize the issue
The AI provides a clear explanation of why the tool call failed based on the trace data.
metorial_get_trace_details
Try it
Copy these to start.
Use these prompts to begin your debugging session.
Starting points
Swap the tool names or error descriptions with your own specific details.
Metorial Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_u9gVDS60iTHUlTLh8tFcQxn710vU1GGlSnGwhKcw/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Metorial capabilities are ready to use.
{
"mcpServers": {
"metorial-mcp": {
"url": "https://edge.vinkius.com/vk_preview_u9gVDS60iTHUlTLh8tFcQxn710vU1GGlSnGwhKcw/mcp"
}
}
}List the most recent execution traces for my agent.
Show me the details for the last failed call in the 'stripe_payment' tool.
Find the trace where the 'search_docs' tool took longer than 3 seconds.
Why did the 'send_email' tool return an error in the last trace?
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Metorial once, then ask.
Link your account once. Your credentials stay encrypted, and you can immediately start asking your agent about execution traces.
Connect Metorial to your AIFAQ
How debugging works with Metorial
- 01
Can the AI fix the tool code for me?
No. The AI only reads the traces to tell you what went wrong. You have to fix the code yourself.
- 02
Can the AI delete my execution traces?
No. The tools only allow the AI to list and retrieve trace details. It cannot delete any logs.
- 03
How does the AI know which trace to look at?
The AI uses the list tool to find relevant traces, then pulls the specific details for the one you ask about.
- 04
Does this work with any MCP client?
Yes. It works with any client that supports the Model Context Protocol, such as Claude, Cursor, or Windsurf.
- 05
Can I see the exact arguments used in a tool call?
Yes. The AI can pull the specific trace details to show you the exact inputs sent to a tool.
More questions about Metorial? The Connector page answers them. See everything the Metorial Connector can do
Connect Metorial to Claude, Cursor, ChatGPT & more
