Logging

GitLab MCP Server uses dual logging — log messages are sent both to stderr (for terminal/file capture) and to the MCP client via the protocol’s logging capability.

Dual output

Destination	Format	Purpose
stderr	Human-readable text or JSON	Terminal output, file redirection, debugging
MCP protocol	`notifications/message`	Displayed in the MCP client’s UI (e.g., VS Code Output panel)

This ensures logs are always visible regardless of the client’s logging support.

Log levels

The server supports four log levels:

Level	When Used
`debug`	Detailed diagnostic information — tool call parameters, API request/response details, session pool operations
`info`	Normal operational events — server startup, tool registration, update checks
`warn`	Non-fatal issues — network timeouts, missing optional configuration, deprecated usage
`error`	Failures — authentication errors, API failures, unrecoverable tool errors

Configuration

Set the log level via environment variable:

# Stdio mode
LOG_LEVEL=debug ./gitlab-mcp-server

# HTTP mode
LOG_LEVEL=info ./gitlab-mcp-server --http --gitlab-url=https://gitlab.example.com

The default level is info.

MCP log messages

When the MCP client supports logging, the server sends structured log notifications:

{
  "jsonrpc": "2.0",
  "method": "notifications/message",
  "params": {
    "level": "info",
    "logger": "gitlab-mcp-server",
    "data": {
      "message": "starting MCP server",
      "transport": "stdio",
      "version": "2.1.0",
      "tools": 40,
      "resources": 24,
      "prompts": 38
    }
  }
}

Security rules

Log messages follow strict security rules:

No tokens — GitLab tokens are never included in log messages
Masked identifiers — In HTTP mode, tokens are shown as ...a1b2 (last 4 characters only)
No PII — User-submitted data is not logged at info level or above
Debug only — Detailed request/response data is only logged at debug level