VClaw

A local-first fleet of AI-controlled machines.
Install a node on each computer you own — then chat with its AI, drive its screen, and run commands from one console, anywhere. Runs on your hardware. You hold the keys and the kill switch.

---

Install

macOS / Linux

curl -fsSL https://get.vclaw.ai/install.sh | sh

Windows (PowerShell)

irm https://get.vclaw.ai/install.ps1 | iex

Not on a custom domain yet? Use the repo directly: curl -fsSL https://raw.githubusercontent.com/vkfolio/vclaw_fleet/main/scripts/install.sh | sh  ·  irm https://raw.githubusercontent.com/vkfolio/vclaw_fleet/main/scripts/install.ps1 | iex

The installer downloads a single, closed-source binary (no Python, no source code), starts the node in the background, auto-starts it at login, and prints your admin URL. Manage it with:

vclaw status      # running? autostart? admin URL
vclaw down        # stop + disable autostart
vclaw up --lan    # expose to other devices on your network
vclaw uninstall   # remove (add --purge to delete all data)

What it is

VClaw turns the computers you already own into a fleet you command from one place:

• One console, every machine — pair each device with a QR code; live status, per-device permissions.
• AI that acts — chat with each machine's agent; it uses real tools and renders interactive controls, under a mode you choose (Ask / Plan / Edit / Auto).
• Skills studio — teach the agent reusable, multi-page playbooks; share them as .vcskill files.
• Remote desktop — Teams-grade WebRTC screen + keyboard/mouse control.
• Quick actions & terminal — one-tap reboot/lock/screenshot/custom commands, plus a live PTY shell.
• Memory, MCP, scheduling — vector recall, pluggable MCP tools, proactive cron/interval tasks.
• Private remote access — reach your fleet from anywhere over a private Cloudflare WARP mesh.

Why it exists

Tools that wire a personal AI through a Telegram/WhatsApp bot or a hosted service make two trades: your data flows through someone else's servers, and your assistant dies when their service does. VClaw removes the middleman — the agent runs on your hardware, reachable privately from anywhere, with you approving anything risky. Read the full story →

Security

• Local-first — your data, keys, and the kill switch stay on your machines.
• Capability-gated — each device gets an allowlist; risky actions wait for your Approve/Deny tap.
• Secrets stay secret — API keys live on the node with locked-down permissions; never sent to the model or to programs it runs.
• Private by default — remote access rides a private mesh or a gated tunnel; nothing public.

Links

• Website / story: https://get.vclaw.ai
• Docs: docs/ — Install · Getting started · Security · FAQ
• Releases: latest binaries

License

VClaw is closed-source. The binaries are free for personal use. This repository hosts the website, docs, and release binaries — not the source. © 2026 Vignesh. All rights reserved.