trace_start

GliderMCP

┌ ┐

└ ┘

│

┌ ┐

[solution]

Starts a long-running command under the workspace and stores an active GliderTrace runtime session.

Why this tool exists

Command and test failures need reproducible execution plus stored stdout, stderr, test, and trace evidence.

How it helps the agent

The model can create durable sessions instead of relying on transient terminal output pasted into a prompt.

Try it locally in the GliderTrace playground

Parameters

Name	Type	Required	Description
command	string	Yes	Command executable to run.
arguments	json	No	Optional command arguments.
workingDirectory	string	No	Optional working directory under the workspace root. Defaults to the workspace root.
environment	json	No	Optional environment variables to set for the command process.
captureStdout	boolean	No	Capture stdout as a sensitive artifact on trace_stop. Defaults to true.
captureStderr	boolean	No	Capture stderr as a sensitive artifact on trace_stop. Defaults to true.
collectTrace	boolean	No	Collect a runtime.nettrace artifact with built-in EventPipe collection. Defaults to true.

Examples

Start a long-running app session

Request 
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "trace_start",
    "arguments": {
      "command": "dotnet",
      "arguments": [
        "run"
      ]
    }
  }
}

Response Notes

Returns active session metadata and process details for a later trace_stop call.

└ ┘