This document explains how the MCP tool-level logging framework handles chained git commands.

When you use git_batch, it creates ONE log entry for the entire batch operation:

• Tool name: git_batch
• Details: "Batch executed: X/Y succeeded"
• Individual commands within the batch are NOT logged separately
• The formatted results show success/failure for each command in the response

1. Cleaner Logs: Avoids cluttering logs with implementation details
2. Atomic Operations: Treats chained operations as single logical units
3. Better Analytics: Easier to track usage patterns of workflows vs individual commands
4. Performance: Reduces log write operations

mcp__git__git_batch(
  repo_path="/path/to/repo",
  commands=[
    {"tool": "git_add", "args": {"files": ["*.py"]}},
    {"tool": "git_commit", "args": {"message": "feat: add feature"}},
    {"tool": "git_push", "args": {"set_upstream": true}}
  ]
)

Instead of three separate log entries:

git_add | STATUS: SUCCESS | Added files: ["file.py"]
git_commit | STATUS: SUCCESS | Committed: abc123
git_push | STATUS: SUCCESS | Pushed main to origin

You get one consolidated entry:

git_batch | STATUS: SUCCESS | Batch executed: 3/3 succeeded

This design follows the principle of abstraction - log at the level of user intent, not implementation details.

All tool calls are logged to ~/mcp-tool-calls.log as configured in the logging framework. Use check-mcp-logs -t to view recent tool calls.