Tailscale's Aperture by Tailscale went prototype to paying customers in months

Tailscale already connects a company's laptops, servers, cloud instances and personal devices into one private network called a tailnet. Aperture takes that same idea and applies it to AI, so companies control model access through the network rather than issuing every employee, agent or tool a separate provider API key.

The build-versus-buy decision is the backdrop here. Tailscale is an infrastructure company, so building the routing and execution layers in-house was the obvious first option, but Remy Guercio says the provider layer looked deceptively simple and was not. Tailscale even shopped the routing layer and the sandbox layer, and it had a working implementation on a different sandbox provider before switching to Vercel — so the choice of AI Gateway and Sandbox reads as a deliberate comparison rather than a default.

Agents then raised the stakes, because an agent that can read private data, act on it and reach the public internet is a security problem now called the 'lethal trifecta.' Aperture's answer is a workflow where a sandbox spins up, Aperture connects to AI Gateway, Tailscale validates identity, the agent works, the sandbox shuts down, and no key is ever issued to the agent. The next test appears to be whether Aperture can hit its stated 'time to first app' goal, which is the outcome the team itself says it must iterate toward.