One sentence, and the whole system works for you

2026-09-03

Clawlink: system-wide control — you give the orders and wait for acceptance

Quick start GitHub

ClawX joins every device into one virtual network, then runs them in parallel with natural language.

Task orchestration and execution

||
|1. Orchestrate tasks (todo.md):|
| - Install the latest Python, pip, and ffmpeg packages|
| - Set up the Remotion video environment|
| - Generate a hello world video under ~/workspace/video and render it to a 4K MP4|
| - Write the tasks above into todo.md|
||
||
|2. Assign it to ComputerA|
| - @ComputerA, finish the tasks in todo.md and update todo.md as each one is done|
||

Devices no longer sit in isolation: Clawlink connects them into one network first, then you orchestrate everything with natural language.

Network + dispatch — one sentence, and it starts
iTunnel — joins all your devices into one virtual LAN
AI Gateway — your natural-language command entry point
One natural-language instruction, and devices get to work — for example: create a hello world Rust project on development machine #5

What you get (before vs now)

From “log in and type commands on every machine” to “tell ClawX one sentence.”

Before: log into each machine and type commands. Now: one sentence via the AI Gateway to the target deviceBefore: log into each machine and type commands. Now: one sentence via the AI Gateway to the target device
Efficiency leap: repeated logins → one order
Before — log into device 1 / 2 / 3 … type commands, repeat n times
Now — tell the Agent one sentence
  • AI Gateway parses intent and issues commands
  • Target devices start executing
You give the orders, then wait for acceptance

Three concrete gains

Three gains: natural-language control, multi-agent parallelism, and issue orders anywhereThree gains: natural-language control, multi-agent parallelism, and issue orders anywhere
Speak plainly · run in parallel · order from anywhere
  • Natural-language control — no commands to memorize; just speak plainly
  • Multi-agent in parallel — A analyzes docs, B gathers news, C checks the network, all at once
  • Issue orders anywhere — one sentence on your phone, and devices follow
The same capability spans status checks, task orchestration, and intel summaries

Use cases

No more logging into every machine, and no more memorizing commands. The three scenarios below all center on the same idea: you say one sentence; Clawlink breaks it down, dispatches it, and runs it in parallel — you only accept the results.

One sentence to query fleet status: AI Gateway fans out in parallel to dev machines, test machines, and edge boxesOne sentence to query fleet status: AI Gateway fans out in parallel to dev machines, test machines, and edge boxes
Scenario 1 · One ask, the whole device fleet in view

On a trip, in a meeting room, or woken by an alert at night — you just need to know: which machines are up, who’s busy, who’s offline. Before, that meant SSH’ing one by one. Now one natural-language sentence on your phone brings back the whole fleet at once.

You: “Check the status of all devices”
AI Gateway parses intent · fans out in parallel
  • Dev machine — online · CPU / RAM / disk at a glance
  • Test machine — online · running tasks A / B / C
  • Edge box — offline · last healthy connection time included
What you get: a device health snapshot anytime
One sentence to assign a todo: draft, execute, and verify across three devices in parallelOne sentence to assign a todo: draft, execute, and verify across three devices in parallel
Scenario 2 · You are the CEO of every terminal

You don’t have to break a todo down onto each machine yourself. State the goal clearly; the Gateway assigns by role: someone drafts, someone executes, someone verifies — multi-agent in parallel, you watch the results.

You: “This todo — each device executes by role”
AI Gateway breaks down the task · targeted dispatch
  • Draft — produce the plan and steps
  • Execute — run the commands and scripts
  • Verify — check results and report back
What you get: one sentence orchestrates a device fleet
Multiple machines gather market intel in parallel and consolidate into one report directoryMultiple machines gather market intel in parallel and consolidate into one report directory
Scenario 3 · Market intel, gathered in parallel across machines

Competitor moves, pricing, and sentiment often live in different environments. Spell out once who does what and where results go; three machines start together, and the report lands in the same directory.

You: “A fetches headlines; B pulls pricing; C searches reviews — all write to D:\report”
AI Gateway splits by device · dispatches in parallel
  • Node A — fetch headline coverage
  • Node B — pull competitor pricing
  • Node C — summarize user reviews
What you get: multi-source intake — open the report and accept

Trust

  • 🔒 Every action is controllable, auditable, and traceable
  • 🏠 Fully on-prem deployment supported; data never leaves the enterprise
  • 🌐 Enter the corporate network securely, with graded permission control

Quick start

Ready to let AI take over your devices?

PowerShell
irm https://irocn.com/download/clawx/install-clawx-win.ps1 | iex
Terminal
curl -fsSL https://irocn.com/download/clawx/install-clawx-linux-x86_64 | sh
Terminal
curl -fsSL https://irocn.com/download/clawx/install-clawx-linux-x86_64 | sh

Tutorial