View MCP Tool Call Logs

Every MCP tool call creates a log entry in the Logs app. Use these entries to confirm which server and tool ran, inspect what the assistant sent, review the response shown to the model, and diagnose failures or latency.

Find an MCP Tool Call

Open the Logs app, then use the most specific filter you have:

Enter the conversation’s Root UUID to see MCP calls from that conversation.
Select the MCP server from the Plugin filter to see calls to that server.
Select mcp.tool.trigger from the Log Type filter to see all MCP tool calls.

You can combine these filters to narrow the results. Individual MCP tools are not available as a filter. Open a log entry to see its tool name.

Read the Log Entry

Each MCP tool call can include the following fields:

Field	What it tells you
Plugin ID	The plugin identifier associated with the log entry.
Execution ID	The identifier for this execution. Include it when escalating a specific call.
MCP server	The server UUID and display name.
Tool name	The MCP tool that the assistant called.
Arguments	The arguments the assistant sent to the tool.
Response for model	The response shown to the model. This may differ from the server’s raw response because of processing, redaction, or truncation.
Execution status	Whether the call succeeded or failed.
Latency	How long the tool call took.
Error message	The recorded error when the tool call fails.

Arguments, the response for the model, and the error message are each limited to 32 KB. A value that exceeds the limit is truncated and ends with ....

Diagnose the Call

Use the log entry to identify where to investigate:

What you see	What to check
No MCP log entry for the conversation	Confirm that the assistant selected an MCP tool. Configure and Test MCP Workspace explains how to review tool selection.
Unexpected or incomplete arguments	Compare the arguments with the user’s request and the tool’s input schema.
A successful call with an unexpected answer	Compare the response shown to the model with the MCP server’s logs and raw response.
A failed call	Review the execution status and error message, then correlate the timestamp with the MCP server’s logs.
High latency or intermittent failures	Check the latency and compare the same time period with upstream server health, timeouts, and rate limits.

For authentication, discovery, selection, timeout, and response-quality guidance, use Limitations and Troubleshooting.

Understand Redacted Values

Standard log redaction masks credentials and common personally identifiable information. Strict Log Redaction can replace log body values with *****. If a value is hidden, review the server’s Strict Log Redaction and audience settings in Launch Configuration. Your organization’s policy may require strict redaction for broadly available servers.

Log Data Security explains the full redaction policy.

Escalate with Useful Evidence

If you cannot resolve the issue, include the following in your escalation:

Root UUID and approximate timestamp with timezone.
MCP server name and tool name.
Execution ID, execution status, and latency.
The sanitized error message and expected behavior.

Do not include access tokens, refresh tokens, credentials, or raw sensitive data.