📊 Token & Quota Impact
Raw terminal SSH drains your model's context window and budget. See how agentic_ssh preserves memory and infrastructure.
| Metric / Scenario | Raw SSH in Prompt | With agentic_ssh MCP | The Difference |
|---|---|---|---|
| Context Consumption | 15,000+ tokens (raw stdout dump) | 185 tokens (structured JSON) | 98.8% token savings 📉 |
| API Cost per Query | ~$0.05 – $0.15 | ~$0.0004 | Fraction of a cent 💰 |
| Security & Blast Radius | Unrestricted ~/.ssh/config access | Strict allow_hosts whitelist | Unauthorized hosts blocked 🛡️ |
| Connection Latency | 800ms – 1.5s handshake reconnect | 0ms (pooled keepalive socket) | Instant execution ⚡️ |
| Long-Running Builds | Blocks agent / silent timeout | Detached async (background: true) | Non-blocking parallel flow 🚀 |
Why AI Agents Struggle with Raw SSH
Standard terminal tools weren't built for Large Language Models. Here's why agents fail over plain SSH—and how agentic_ssh solves it:
The Context Avalanche
When an agent runs a build or package upgrade, 10,000 lines of GCC/Cargo compiler noise dump straight into your LLM context window.agentic_ssh auto-summarizes output and redirects full streams to clean session logs.
Silent Death by Dropout
Tailscale, firewalls, and cloud NATs love to silently drop idle sockets during long compiles.agentic_ssh uses Rust-native keepalives and automatic connection pooling that never sleeps.
Nested Escaping Hell
Asking an LLM to quote bash inside SSH inside an MCP tool string leads to constant syntax errors.agentic_ssh handles arguments and command joining natively behind the scenes.
Safe-by-Default Whitelists
A hallucinating agent shouldn't touch your production database.agentic_ssh enforces strict host boundaries (allow_hosts) and cryptographic local config signatures.
🤖 Copy & Paste Agent Prompt Recipes
Drop these directly into Claude Code, Cursor, Gemini, or Antigravity to see agentic_ssh in action:
"Check the CPU load, RAM usage, and available disk space across stan, cartman, and aruba. Report any bottlenecks."
"Inspect all listening TCP and UDP ports on cartman. Flag any processes bound to 0.0.0.0 that should be private."
"Check if any Docker containers crashed on stan, and tail the last 40 lines of the web container logs."
"Run the database backup on aruba in the background and notify me when the session log indicates success."
🖥️ Built-in CLI Tools for Humans
agentic_ssh is also a full-featured terminal CLI for developers.
alias ash="agentic_ssh" to your ~/.zshrc or ~/.bashrc!
Multi-host streaming TUI grid across multiple servers.
agentic_ssh watch fleet 'uptime'Direct JSON queries from terminal for shell scripting.
agentic_ssh json list_ports stanInstant diagnostic scan of keys and agent configs.
agentic_ssh doctor⚡️ Installation & Setup
Install Binary
Install via Homebrew, Cargo, or Nix:
brew install sandbanks/tap/agentic_sshAuto-Register MCP
Auto-configures your installed AI agents:
agentic_ssh installRun Doctor
Verify SSH keys and agent connections:
agentic_ssh doctor🧰 Built-in MCP Tools Catalog
| Tool Name | Description | Key Arguments |
|---|---|---|
| list_hosts | Lists all discovered and permitted SSH hosts from ~/.ssh/config. | none |
| run_command | Executes shell commands across hosts concurrently (supports async background mode). | hosts, command, background |
| get_system_stats | Retrieves CPU load, RAM usage, and disk partition stats. | hosts |
| list_ports | Scans listening TCP/UDP ports with PID & binary attribution. | hosts, protocol |
| check_docker_status | Inspects Docker daemon health and container runtime metrics. | hosts |
| tail_container_logs | Tails recent frames from target Docker container logs. | hosts, container, lines |
| search_processes | Searches and evaluates active processes via regex. | hosts, pattern |
Ready to give your AI agent SSH superpowers?
Install with one command and start exploring your fleet safely.