feat: PM2 Native MCP Server — WORKING EXAMPLE

PM2 should optionally include a native MCP server to expose process introspection, logs, and control via HTTP/SSE.

Enablement:

pm2 start ecosystem.config.js --mcp
# or
pm2 set mcp.enabled true
pm2 set mcp.port 7777

Endpoints / Features:

/processes (GET) — list all processes, status, CPU/mem, uptime
/process/:id (GET) — detailed info on a process
/logs/:id (GET) — stream or search logs for a process (SSE or tail)
/restart/:id (POST) — restart a process
/stop/:id (POST) — stop a process
/scale/:id (POST) — scale cluster instances up/down
/metrics (GET) — CPU/mem, restart counts, uptime summary
/health (GET) — overall PM2 daemon health

Design Notes:

• MCP server disabled by default
• Authentication optional (token or local-only by default)
• Can serve multiple MCP clients concurrently
• Minimal footprint — reuses PM2 internal process table and log streams

Use Cases:

LLM / AI tool orchestration (query, restart, scale)
Dev workflows: auto-tail logs, search history
Ops dashboards / monitoring integration
Programmatic PM2 control without shell commands

Summary:

• Native, optional MCP server
• Exposes logs, metrics, status, and control
• Lightweight, secure, opt-in

Start by tracing PM2's existing CLI configuration and daemon interfaces, especially the internal process table and log streams mentioned in the issue. Define the integration points for the optional HTTP/SSE server and verify that the documented process, log, control, metrics, and health endpoints work while the server remains disabled by default.