meshpop 4.13.53

MeshPOP Fleet Orchestration — manage mesh VPN servers at scale

meshpop (mpop)

All-in-one server fleet management — monitor, manage, and automate multi-server infrastructure through AI.

pip install meshpop

No external dependencies. Pure Python. Installs in seconds on any Linux or macOS machine.

---

Why mpop?

The problem with managing server fleets today

Running more than a handful of servers means constantly jumping between machines, tools, and dashboards:

# Old way: checking 10 servers one by one
ssh root@web1 "df -h && free -h && uptime"
ssh root@web2 "df -h && free -h && uptime"
ssh root@db1  "df -h && free -h && uptime"
# ... repeat 7 more times
# No unified view. No anomaly detection. Pure manual work.

Monitoring tools (Prometheus, Grafana, Zabbix) solve the visibility problem but require:

• Dedicated monitoring servers
• Agent installation and configuration on every node
• Dashboard setup and maintenance
• Learning curve for each tool

Configuration management (Ansible, Chef, Puppet) automates deployments but:

• Requires a separate control node
• YAML/DSL to learn
• No real-time monitoring built in

mpop is a single command that does all of this — and lets an AI manage it for you.

Before mpop

# Checking fleet health — manual SSH to each server
for server in web1 web2 web3 db1 db2 worker1 worker2; do
  echo "=== $server ==="
  ssh root@$server "
    echo 'Disk:' && df -h / | tail -1
    echo 'Memory:' && free -h | grep Mem
    echo 'Load:' && uptime
    echo 'Failed services:' && systemctl --failed --no-pager
    echo 'Security:' && last | head -5
  "
done
# 5-10 minutes of work
# Still no unified anomaly detection
# Still have to interpret the output yourself

After mpop

mpop              # Full fleet dashboard in 3 seconds
mpop heal         # Auto-detect and fix issues on all servers
mpop ask "which server is running low on disk?"

Or let AI do it — just describe what you want in plain language.

Comparison

	Manual SSH	Monitoring Stack	mpop
Fleet overview	One server at a time	After setup + config	mpop — instant
Setup time	None	Days to weeks	Minutes
Dependencies	SSH keys	Multiple services	None
Anomaly detection	Manual inspection	After threshold config	Built-in AI analysis
Fix issues	Manual commands	Alerts only	mpop heal — auto
Natural language queries	✗	✗	✓ mpop ask "..."
AI management (MCP)	✗	✗	✓ Full MCP integration
Agent deployment	Manual	Manual	mpop setup <server>

---

Architecture

mpop works as a three-layer system:

┌─────────────────────────────────────────────────────────────┐
│  mpop CLI / MCP  ← your machine (or AI agent)               │
│  Talks to all servers simultaneously                         │
└──────────────────────────┬──────────────────────────────────┘
                           │  HTTP :8721
           ┌───────────────┼───────────────┐
           │               │               │
    ┌──────▼──────┐  ┌──────▼──────┐  ┌──────▼──────┐
    │  agent on   │  │  agent on   │  │  agent on   │
    │   web1      │  │   web2      │  │   db1       │
    │  :8721      │  │  :8721      │  │  :8721      │
    └─────────────┘  └─────────────┘  └─────────────┘
         Gossip protocol — agents share stats with each other
         every 30 seconds. Any agent knows the full fleet state.

Each server runs a lightweight agent (agent_reporter.py) on port 8721. Agents collect local stats and gossip them to peer agents. mpop queries any one agent and gets the full fleet picture.

mpop is Layer 3 of the MeshPOP stack:

Layer 3  mpop    Fleet orchestration — monitor, manage, automate  ← this
Layer 2  vssh    Authenticated transport — remote exec, file transfer
Layer 1  wire    Encrypted mesh VPN — connects all nodes

Each layer is independently installable. mpop works without wire or vssh — it can reach agents over any network (LAN, VPN, public internet).

---

Installation

Install mpop

pip install meshpop

This installs:

• mpop — CLI command
• mpop-mcp — MCP server for AI integration

No external packages required. Pure Python standard library.

Configure servers

mpop needs to know which servers to manage. Create ~/.mpop/config.json:

{
  "node": "myserver",
  "node_name": "myserver",
  "dashboard_url": "http://10.0.0.1:8721",
  "servers": {
    "web1":    {"ip": "10.0.0.10"},
    "web2":    {"ip": "10.0.0.11"},
    "db1":     {"ip": "10.0.0.20"},
    "storage": {"ip": "10.0.0.30"}
  },
  "hostname_map": {
    "web-server-1": "web1",
    "web-server-2": "web2"
  }
}

hostname_map is optional — use it to normalize agent-reported hostnames to short canonical names.

Deploy the agent to each server

The agent collects stats (CPU, memory, disk, GPU, processes, services) and serves them on port 8721.

# Install and start the agent on a remote server:
mpop setup web1

# Or deploy to multiple servers:
mpop setup web1 && mpop setup web2 && mpop setup db1

# Check that agents are running:
mpop servers

The agent runs as a lightweight Python process — typically ~10MB RAM, negligible CPU.

Start monitoring

mpop          # Full dashboard — all servers at a glance

mpop dashboard  v4.13.52
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  NODE   STATUS   CPU   MEM    DISK    GPU    UPTIME
  web1   ONLINE   12%   45%    38%     —      14d 3h
  web2   ONLINE    8%   32%    41%     —      14d 3h
  db1    ONLINE   23%   67%    55%     —       7d 12h
  gpu1   ONLINE   91%   78%    44%    94%      2d 6h
  ○ backup  OFFLINE  —     —      —       —      —
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  5 servers  4 online  1 offline   14:23:07

---

CLI Reference

Dashboard and monitoring

mpop                       # Main dashboard
mpop servers               # List all servers with IPs and roles
mpop info web1             # Deep dive — GPU, processes, security, agent
mpop full                  # Comprehensive 9-table report with AI summary
mpop hw                    # Local hardware — CPU, GPU, RAM, disk, NIC
mpop temp                  # CPU/GPU temperatures across fleet
mpop gpu                   # GPU status — VRAM, utilization, processes
mpop services              # Running services on each server
mpop logs web1 auth        # View logs (auth / nginx / system)
mpop watch                 # Real-time monitoring with anomaly alerts
mpop trend                 # Historical trends — memory, disk, uptime
mpop matrix                # Server-to-server connectivity matrix
mpop diff web1 web2        # Compare two servers side-by-side

---

Security

mpop security              # Security overview — firewall, fail2ban, SSH
mpop audit web1            # Detailed security audit of one server
mpop ssh-attacks           # SSH attack analysis — top IPs, geo distribution

---

Network and VPN

mpop vpn                   # VPN status (WireGuard / Tailscale)
mpop peers                 # Connected VPN peers and IPs
mpop watchdog start        # VPN watchdog — auto-restart on disconnect

---

Remote execution

mpop exec web1 "df -h"             # Run command on a server
mpop exec web1 "systemctl restart nginx"
mpop python web1 "import os; print(os.uname())"   # Run Python code

---

File operations

mpop read-file web1 /etc/nginx/nginx.conf
mpop scp ./config.json web1 /etc/app/config.json

---

NAS operations

mpop nas-ls /backups
mpop nas-read /backups/latest.log
mpop nas-backup web1 /var/data /backups/web1
mpop nas-usage /backups

---

AI diagnostics

mpop heal                  # Auto-detect problems and suggest fixes
mpop heal web1 --execute   # Auto-detect AND apply fixes
mpop predict web1          # Predict future issues (disk full, memory pressure)
mpop ai                    # AI trend analysis across fleet
mpop advise                # Infrastructure optimization recommendations
mpop logai                 # AI-powered cross-server log analysis
mpop ask "which server is low on disk?"   # Natural language → mpop command
mpop ask "restart nginx on all web servers" --execute

---

Deployment and config

mpop setup web1            # Install agent on remote server
mpop deploy web1           # Deploy mpop binary to a server
mpop config show           # View current config
mpop config set key value  # Modify config
mpop backup web1           # Backup server configurations

---

Agents and automation

# Create a custom automation agent from a description:
mpop agent-make --name disk-watcher \
  --behavior "Alert when any server disk exceeds 80%"

# Deploy it to servers:
mpop agent-deploy disk-watcher web1 web2 db1

# Run it:
mpop agent-run disk-watcher

# Chain multiple agents (pipe output between them):
mpop agent-chain run my-workflow

---

Knowledge and notifications

mpop brain search "nginx config"        # Search knowledge base
mpop notify slack "Deploy complete"     # Send Slack / Discord / Telegram message
mpop export                             # Generate shareable HTML status report

---

AI Management via MCP

mpop ships with an MCP server that lets AI agents (Claude, etc.) manage your entire fleet through natural language — from the first server setup through ongoing operations and incident response.

Setup

pip install meshpop   # installs mpop-mcp automatically

Add to Claude config (~/.claude/settings.json):

{
  "mcpServers": {
    "mpop": { "command": "mpop-mcp" }
  }
}

If mpop-mcp is not on PATH:

{
  "mcpServers": {
    "mpop": {
      "command": "python3",
      "args": ["-m", "mpop_mcp_server"]
    }
  }
}

What the AI can do

Once connected, you can describe what you want in plain language:

"Show me the fleet dashboard" "Which servers are running low on disk?" "Restart nginx on all web servers" "What's using GPU memory on gpu1?" "Run a security audit on db1 and tell me what you find" "Set up the monitoring agent on the new server at 10.0.0.50" "Are there any SSH attack patterns I should know about?" "Predict which servers might have issues in the next week"

The AI handles everything — it runs the right mpop commands, interprets the output, and gives you a clear answer.

From zero to monitored fleet with AI

You: "I have 3 new servers at 10.0.0.10, 10.0.0.11, 10.0.0.20.
     Set them up for monitoring as web1, web2, and db1."

AI:
1. Updates ~/.mpop/config.json with the three servers
2. mpop_setup(web1) — deploys agent
3. mpop_setup(web2) — deploys agent
4. mpop_setup(db1)  — deploys agent
5. mpop_servers     — verifies all three are online
6. mpop_dashboard   — shows the fleet dashboard
7. Reports: "All three servers are online and being monitored.
             web1: 45% disk, 32% mem. web2: 41% disk, 28% mem.
             db1: 55% disk, 67% mem — you may want to watch that."

MCP Tools Reference

Overview and monitoring

Tool	Description
mpop_dashboard	Main fleet dashboard
mpop_servers	List all servers with names, IPs, roles
mpop_info	Deep server details — GPU, processes, security, agent
mpop_full	Comprehensive 9-table report with AI summary
mpop_hw	Local hardware — CPU, GPU, RAM, disk, NIC
mpop_temp	CPU/GPU temperatures across servers
mpop_gpu	GPU status — VRAM, utilization, processes
mpop_services	Running services on each server
mpop_logs	View server logs (auth, nginx, system)
mpop_watch	Real-time monitoring with anomaly alerts
mpop_trend	Historical trends — memory, disk, uptime
mpop_matrix	Server-to-server connectivity matrix
mpop_diff	Compare two servers side-by-side

Security

Tool	Description
mpop_security	Security overview — firewall, fail2ban, ports, SSH
mpop_audit	Detailed security audit of a server
mpop_ssh_attacks	SSH attack analysis — patterns, top IPs, geo

Network and VPN

Tool	Description
mpop_vpn	VPN status (WireGuard + Tailscale)
mpop_peers	Connected VPN peers
mpop_watchdog	VPN watchdog daemon — auto-restart on disconnect

Execution

Tool	Description
mpop_exec	Run shell command on any server
mpop_python	Run Python code on local or remote server
mpop_raw	Run raw mpop CLI command with pipes/redirects
mpop_delete	Safe file deletion (system paths blocked)

File operations

Tool	Description
mpop_read_file	Read file (local or remote)
mpop_write_file	Write to local file
mpop_scp	Transfer file to/from server

NAS

Tool	Description
mpop_nas_ls	List NAS files
mpop_nas_read	Read NAS file
mpop_nas_write	Write to NAS
mpop_nas_backup	Backup server data to NAS
mpop_nas_usage	NAS disk usage summary

AI and diagnostics

Tool	Description
mpop_ai	AI trend analysis — predicts status changes
mpop_heal	Auto-detect problems and suggest/apply fixes
mpop_predict	Predict future issues (disk full, memory pressure)
mpop_advise	Infrastructure optimization advisor
mpop_logai	AI-powered cross-server log analysis
mpop_ask	Natural language → mpop commands

Config and deploy

Tool	Description
mpop_config	View/modify mpop configuration
mpop_setup	Install mpop agent on remote server
mpop_deploy	Deploy files or binary to servers
mpop_backup	Backup server configurations
mpop_secret	Encrypted credential management

Agents and automation

Tool	Description
mpop_agent_list	List custom automation agents
mpop_agent_make	Create agent from description
mpop_agent_deploy	Deploy agent to servers
mpop_agent_run	Run an agent
mpop_agent_workflow	Multi-agent sequential workflows
mpop_agent_chain	Pipe agent outputs via JSON

Knowledge and integration

Tool	Description
mpop_brain	Search knowledge base
mpop_project	Project management — status, history
mpop_memory	Code knowledge base + Q&A
mpop_rag	Document Q&A with RAG
mpop_notify	Send via Slack / Discord / Telegram / Email
mpop_webhook	Webhook management
mpop_workflow	Predefined workflows (health, security, deploy)
mpop_export	Generate shareable HTML status report

---

Requirements

• Python 3.8+
• Servers accessible via SSH or VPN (wire, Tailscale, or any network)
• psutil on each monitored server (for memory/disk stats) — pip install psutil

---

Links

• Wire VPN: github.com/meshpop/wire
• vssh transport: github.com/meshpop/vssh
• PyPI: pypi.org/project/meshpop

License

MIT — MeshPOP