Login
This will change how you homelab | Aperture GA announcement
Getting an AI agent to SSH into your server is easy. Giving it only the access it actually needs is the interesting part.
Aperture by Tailscale is now generally available, and it can give agents controlled SSH access to machines on your tailnet. In this Summer Update, Remy shows us how Aperture can set up real homelab services while keeping Tailscale identity and access controls in the loop.
We use a Raspberry Pi to see how an agent can install and configure services, add them to a tailnet, and work within the permissions you define. We also look at Aperture's mobile experience, LLM and MCP visibility, and how its SSH and node-management capabilities can be used from other agent workflows through MCP.
The goal isn't simply to let an agent onto your network. It's to make intentional decisions about what that agent can and can't access.
Learn more about Aperture: http://tailscale.com/blog/aperture-ga
Get started with Aperture: https://tailscale.com/docs/aperture/get-started
Watch the Tailscale Summer Update playlist: https://www.youtube.com/playlist?list=PLJW-_hiaseM8
In this video:
00:00 Why homelabbing is still too complicated
00:24 Aperture is now generally available
01:00 From LLM proxy to AI gateway
01:31 Why not just give an agent an SSH key?
02:03 Putting an agent to work in the homelab
02:57 Giving services their own Tailscale identity
03:12 Taking Aperture mobile
03:37 Seeing what your agents are doing
04:07 Using Aperture with other agents through MCP
04:34 Why the guardrails matter
05:03 Bringing Aperture tools to your existing workflow
05:15 Try Aperture
