How a Tailscale connection, and Aperture, can reduce some (but not all) of the blind spots with agentic tools like OpenClaw.

Tailscale is a platform providing secure networking solutions for teams and businesses. Readers can learn about VPN technology, network security, and remote access solutions. With articles, tutorials, and case studies, Tailscale offers  insights and expertise for implementing secure networking infrastructure and protecting sensitive data.

Tailscale

Tailscale explains how to run AI agents like OpenClaw more safely using two approaches: routing LLM API calls through Aperture (their AI gateway) to prevent API key leakage and gain cost/usage visibility, and using Tailscale Serve instead of Funnel to keep the agent private to your tailnet. Aperture removes the need to store sensitive credentials on the agent's machine, logs tool calls and token usage, and supports spending caps. The post also warns strongly against exposing OpenClaw to the public internet via Funnel, and shares a real experiment running OpenClaw as a news-gathering agent with costs averaging $7/day.

What Tailscale can (and can't) fix with OpenClaw

Seeing tool calls and setting cost quotas: an experiment

Connecting to danger (please don’t Funnel your OpenClaw)