Your servers,
one command away.
Monitor, deploy, connect, and search across your entire fleet. Mesh VPN, instant SSH, distributed code search — all from one CLI.
curl -fsSL mpop.dev/install | bash
Linux & macOS · No root required · ~30 seconds
Toolchain
Five integrated tools that communicate through the Wire mesh VPN.
| Tool | Description | Type | Install |
|---|---|---|---|
mpop Fleet Manager | Monitor CPU, memory, GPU, services, logs, and run security audits across all servers. | CLI | $ pip install meshpop |
Wire Mesh Network | WireGuard-based mesh VPN with NAT traversal, auto-discovery, and relay fallback. | VPN | $ mpop.dev/wire |
vssh Secure Shell | High-speed SSH & file transfer over mesh. 50+ MB/s, P2P tunnels, port forwarding. | SSH | $ mpop.dev/vssh |
MeshDB Code Search | Distributed full-text code search with BM25 ranking and semantic embeddings. | Search | $ mpop.dev/meshdb |
Agent Remote Exec | Lightweight agent for metrics collection, health checks, and remote command execution. | Agent | $ mpop.dev/agent |
Architecture
All tools communicate through the Wire mesh VPN. No public IP required — every server is one hop away.
Zero-config VPN
WireGuard mesh with NAT traversal. Connect any machine in seconds.
Instant fleet search
Full-text code search across all servers in under 1 second.
50+ MB/s transfers
P2P file transfer through the mesh. No speed limits.
Auto-healing agents
AI-powered diagnostics and predictions.
Unified CLI for server fleet monitoring, security auditing, AI diagnostics, and remote execution.
Installation
Requires Python 3.8+. Compatible with Linux, macOS, and WSL.
Getting Started
After installation, run mpop config auto to auto-detect servers on your Wire VPN mesh.
Key Commands
| Command | Description |
|---|---|
| mpop dash | Dashboard — CPU, memory, disk, uptime for all servers |
| mpop info [server] | Detailed info — GPU, processes, security status |
| mpop exec "cmd" | Run shell command on all or specific servers |
| mpop services | Check running services (nginx, docker, ollama, etc.) |
| mpop logs [server] | View server logs — auth, nginx, system |
| mpop security | Security overview — firewall, fail2ban, open ports |
| mpop audit [server] | Deep security audit — SSH attack analysis |
| mpop gpu | GPU status — VRAM, utilization, processes |
| mpop temp | CPU/GPU temperatures across all servers |
| mpop matrix | Server-to-server connectivity test |
| mpop diff s1 s2 | Side-by-side comparison of two servers |
| mpop trend | Historical resource trends (24h) |
| mpop heal [server] | AI auto-diagnosis and fix suggestions |
| mpop predict [server] | Predict future issues — disk full, memory pressure |
| mpop notify slack "msg" | Notification via Slack, Discord, Telegram, email |
MCP Integration (Claude Code)
mpop includes an MCP server that lets Claude Code directly manage your servers.
Once configured, Claude can check server health, run remote commands, diagnose issues, and manage agents through natural conversation.
WireGuard-based mesh VPN with NAT traversal, auto-discovery, and relay fallback. The network backbone of the entire mpop ecosystem.
Installation
Why VPS over Laptop?
Wire works on any machine — a laptop, Raspberry Pi, or VPS. But for a reliable mesh network, a VPS provides critical advantages:
- Always-on connectivity — Laptops sleep, lose Wi-Fi, or get closed. A VPS runs 24/7, so other nodes can always find each other.
- Stable public IP — NAT traversal works most of the time, but some networks (corporate firewalls, CGNAT) block P2P connections. A VPS with a public IP acts as a reliable relay.
- Consistent latency — VPS providers offer low-latency, high-bandwidth connections. Laptop connections vary with Wi-Fi and ISP quality.
- Central coordination — Wire's discovery server runs best on a VPS. All peers register through it, and the VPS acts as the seed node.
- Cost is minimal — A $5/month VPS (Vultr, Hetzner, DigitalOcean) is enough. Wire uses less than 50MB of RAM.
How It Works
Wire creates a WireGuard mesh where every node can reach every other node. Each node gets a private VPN IP (e.g., 10.99.x.x). The coordination server distributes public keys and endpoints.
NAT Traversal
Wire uses UDP hole-punching for direct P2P connections, even behind NAT. If direct connection fails (symmetric NAT, strict firewall), traffic routes through the relay node automatically.
Watchdog
Built-in watchdog checks VPN connectivity every 5 minutes. If the connection drops, it re-establishes automatically. Install with wire watchdog install.
Key Commands
| Command | Description |
|---|---|
| wire status | VPN status — running state, connected peers, network info |
| wire peers | List all connected peers with endpoints |
| wire connect URL | Join a Wire mesh network |
| wire add-node NAME | Add a new node to the network |
| wire remove-node NAME | Remove a node from the network |
| wire watchdog | Auto-recovery watchdog status |
| wire diagnose | Diagnose connectivity issues |
MCP Integration (Claude Code)
Claude can check VPN status, list peers, diagnose issues, and manage nodes through natural conversation.
High-speed SSH and file transfer over the Wire mesh. 50+ MB/s, P2P tunnels, port forwarding — without exposing SSH to the public internet.
Installation
vssh requires Wire to be installed and connected. All connections go through the VPN mesh.
Why vssh?
- No public SSH exposure — All connections through the Wire mesh. Port 22 stays closed to the internet.
- Name-based access — Connect by server name (
g1,v1) instead of IPs. - 50+ MB/s transfers — P2P through WireGuard, bypassing traditional SSH bottlenecks.
- Built-in sync — Sync files between any two servers without rsync or SCP config.
Key Commands
| Command | Description |
|---|---|
| vssh status | Connection status to all servers with latency |
| vssh exec SERVER "cmd" | Execute a command on a remote server |
| vssh put SERVER local remote | Upload a file (50+ MB/s) |
| vssh get SERVER remote local | Download a file from a server |
| vssh sync SRC DST path | Sync files between two servers |
| vssh tunnel SERVER LP RP | Create SSH tunnel (port forwarding) |
| vssh speed-test SERVER | Test transfer speed |
| vssh p2p-status | Check P2P NAT hole-punch capability |
| vssh keys | Manage SSH keys and secrets |
Usage Examples
MCP Integration (Claude Code)
Claude can execute remote commands, transfer files, create tunnels, and check connectivity through vssh.
Distributed full-text code search across all servers. BM25 ranking, semantic embeddings, parallel execution — typically under 500ms.
Installation
Each server maintains its own local index. Searches query all servers in parallel through the Wire mesh.
Key Commands
| Command | Description |
|---|---|
| meshdb search "query" | Full-text search across all servers (BM25) |
| meshdb find "filename" | Find files by name across all servers |
| meshdb read SERVER path | Read indexed file contents |
| meshdb semantic "query" | AI semantic search via ChromaDB |
| meshdb index PATH | Index a directory (local or remote) |
| meshdb status | Index status — total docs, types, top directories |
Search Syntax
MeshDB uses FTS5 (SQLite full-text search) with powerful operators:
Semantic Search
Find code by meaning, not just keywords. Uses ChromaDB embeddings for intelligent matching.
MCP Integration (Claude Code)
Claude can search code, find files, read contents, and perform semantic search across your fleet.
Lightweight daemon for metrics collection, health checks, and remote command execution. Runs on each node to provide real-time status to the mpop control plane.
Installation
The agent runs as a systemd service and communicates over the Wire mesh VPN.
What the Agent Does
- Metrics collection — CPU, memory, disk, GPU, temperatures, processes. Reports every 30 seconds.
- Health checks — Monitors critical services, disk space, detects anomalies.
- Remote execution — Accepts commands from mpop control plane, authenticated and sandboxed.
- Auto-healing — Restarts failed services, cleans temp files, rotates logs.
- Security monitoring — Watches auth logs for brute force, reports firewall and SSH status.
Custom Agents
Create custom automation agents with AI-generated code:
Agent Lifecycle
Agents can be triggered manually, scheduled via cron, or chained into multi-step workflows. Each agent produces JSON output that pipes into the next agent in a chain.
MCP Integration (Claude Code)
Claude can create agents from natural language, deploy them, run them, and chain them into workflows.