You built an AI memory system. Now your agent needs hands. Here are 6 extensions that compound — from household knowledge to the job hunt.

“I built it. Now what?”

That’s been the most common message since the Open Brain guide shipped. People aren’t writing in with bug reports or setup questions. They’re staring at a working personal AI memory system and having no idea what to do next.

I know the feeling. You followed the guide, the database is live, your agent reads from it and writes to it and remembers what you said last Tuesday. The infrastructure works. And now you’re sitting there realizing that the hard part was never the build. The hard part is knowing what to put in it.

The usual advice doesn’t help much. “Capture your thoughts.” “Ask it questions.” That’s like handing someone who just built a workshop full of beautiful tools a piece of sandpaper and telling them to get sanding.

Here’s what changed for me: I started building extensions on Open Brain where both I and my agent could see and act on the same data. A maintenance log where my agent catches that a warranty is about to expire based on something a technician mentioned eighteen months ago, and I pull up the same log on my phone when the repair tech asks what was done last time. A family schedule where my agent cross-references both parents’ calendars and every kid’s activity simultaneously, and Sunday night we’re both looking at the same view on a tablet planning the week. A job search dashboard where my agent spots that a warm introduction is going cold while I’m drowning in eleven other workstreams, and I see the whole pipeline at a glance over coffee.

The pattern underneath all of these is the same: a shared surface with two doors. Your agent enters through one. You enter through the other. Both sides read the same data, both sides write to it, and each one does what it’s best at.

Here’s what’s inside:

• The two-door principle. Why every Open Brain extension needs an agent door and a human door, and why a chat window alone will always be a keyhole into your own data.

• Six use cases from household knowledge to the job hunt. Each one built around the problems that actually defeat people, with specific examples of how conversational AI and autonomous agents handle them differently.

• Four design principles that generate your own use cases. Time-bridging, cross-category reasoning, proactive surfacing, and the judgment line that determines whether you trust the system or abandon it.

• The pull/push paradigm. Why Claude, ChatGPT, and OpenClaw are three different interfaces to the same database, and understanding when to use each one changes everything.

• The full build guide and companion prompts. An open-source repo with six extensions you build in order, companion prompts, and a community contribution system — everything you need to go from principles to working tables.

Let me show you what this looks like when it’s running.