Private beta — join the waitlist

Secure access to your
AI machines from anywhere

SSH into your GPU box, open Ollama from your phone, and share ComfyUI with one link — without exposing your home network.

No credit card required · Free tier available

Three demos. That's the product.

📱

Open my Ollama from my phone

Sign in on mobile, tap your GX10, chat with Llama 3. Tokens stream from your home GPU over LTE.

$ anyfy up
✓ GX10 online → open app
🎨

Share my ComfyUI with a friend in one link

Toggle "Share" on your ComfyUI tunnel. Copy the link. Your friend opens it without installing anything.

https://xyz.tun.anyfyio.com
← your ComfyUI, shared
💻

SSH into my GX10 from the browser

Open the dashboard, click Terminal. Full SSH session in your browser, no VPN, no port forwarding.

$ anyfy ssh gx10
user@gx10:~$

Why agent-only?

Mesh VPNs are powerful — but they put a virtual network on every device, which is more than most AI builders need. Anyfy takes the reverse-tunnel approach: your agent connects out, nothing opens in.

  • Zero open ports. No router changes, no firewall rules. Your machine stays invisible.
  • Zero mesh. No virtual network adapter, no kernel networking, no shared IP space between your devices.
  • One process. A single lightweight binary. Start it, forget it. Your services show up in the dashboard.
See full comparison vs Tailscale, ngrok, Cloudflare →
# On your GPU box
$ curl -fsSL anyfyio.com/install.sh | sh
✓ anyfy installed
$ anyfy login
✓ authenticated as you@company.com
$ anyfy up
✓ GX10 online
✓ Ollama detected on :11434
Open your phone. Tap GX10. Start chatting.

How does Anyfy compare?

Compared to Tailscale, ngrok, and Cloudflare Tunnel. Full comparison →

Feature Anyfy ← you Tailscale ngrok Cloudflare Tunnel
AI-native UX (chat, previews)
Built-in Ollama chat UI
Mobile chat client
OpenAI-compatible proxy Aperture (paid)
Mesh VPN required No Yes No No
Open ports on device None None None None
Setup time ~60s ~5 min ~2 min ~10 min
Free tier devices 1 100 1 tunnel 1 tunnel
SSH with session recording ✓ (Team+) ✓ (Business)
Per-share API keys

How it works

One agent. One command. Everything connected.

1

Install the agent

Run one command on your GPU box. The agent connects to the nearest relay.

2

See it online

Your machine appears in the dashboard in under 2 seconds.

3

Access from anywhere

SSH, Ollama, ComfyUI, Jupyter — available from any device, anywhere.

Frequently asked questions

Why not just use Tailscale?

Tailscale is excellent for full network mesh — every device on a shared virtual LAN. That's powerful but more than most AI builders need, and it puts a virtual network adapter on every machine. Anyfy is agent-only (no kernel networking, zero open ports), ships a built-in chat UI against your Ollama, and has a mobile-first design for consuming AI, not just administering machines. If you want full mesh, use Tailscale. If you want to open your Ollama from your phone in 60 seconds, use Anyfy.

Why not ngrok?

ngrok is built for exposing development servers to the web. It's great for webhooks and demos. Anyfy is built for persistent, private access to AI machines — with per-device identity, org-level ACL, SSH brokering, and an Ollama chat UI. ngrok has no concept of a 'device' or 'org.'

Is this a VPN?

No. Anyfy does not create a virtual network between your machines. It is a reverse-tunnel service: your agent connects out to Anyfy's relay, and the relay securely forwards specific traffic back to specific local services. Nothing else on the network changes. Your machines remain invisible to the internet.

Do I need to open ports on my router?

No. The agent initiates the connection outbound. No inbound ports required, no router configuration, no firewall rules.

What's the difference between a 'Share' and 'Reach'?

'Share' creates a public link anyone can open — useful for sharing a ComfyUI session with a collaborator. 'Reach' creates a private hostname inside your org, visible only to signed-in members — useful for team-internal Jupyter notebooks or databases.

Ready to connect your GPU?

Join the waitlist and be first to get access.