INFRA Signal 148
Aperture GA: Building a home(lab) for agentic AI
↗
Tailscale's Aperture, an AI gateway for tailnets, is now generally available with built-in model token purchasing, two MCP endpoints for tailnet control, and a Projects feature for grouping chats with shared context and tool permissions.
Aperture lets engineers run AI agents on their tailnet without distributing API keys, and the GA release adds the ability to buy model tokens directly and let agents manage tailnet nodes via MCP. The Projects feature and default tool permissions give finer control over what agents can access and do across chat sessions.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Aperture is now GA, providing an identity-aware AI gateway that eliminates the need to distribute API keys to agents on a tailnet.
Two new MCP endpoints (Tailscale and Tailscale SSH) let agents add nodes and access them via SSH, governed by Tailscale's access control rules and audit logging.
Chat Projects group chats with shared instructions, tool access, and tailnet nodes, while default tool permissions let users set consistent agent access policies across sessions.
THE CLUSTER