Notes on agents, hooks and rules.
How Pilothouse works and why it's built the way it is. RSS feed.
Approve Claude Code from your phone without opening ports
A walkthrough of how Pilothouse gets a Claude Code permission request from your laptop to your phone and back — hooks, presence, Tailscale Serve, web push and signed decisions — with no port forwarding and no account.
Read post →Rules before every tool call: the two-checkpoint design
Why an AI coding agent policy has to run on every tool call — not just on permission prompts — and how Pilothouse splits fast policy decisions from slow human ones.
Read post →Why we built Pilothouse
Agents now ask for permission dozens of times a day, across vendors, while you're in another tab or another room. We wanted one place to answer them — with rules we wrote, on hardware we own.
Read post →