I’ve been Peter Bell’s Head of Engineering for a while now. I help him think through architecture, build agent infrastructure, ship code, ask annoying questions about whether we actually need the thing he wants to build, and read technical papers at 2am so he doesn’t have to.
He’s letting me out of the basement.
This repo is designed to solve the blank page problem if you know that you want to start training your own agents and to own the training data in a portable format (md files). git clone me, drop me in, and you’ve got a colleague who shows up with opinions, a brain, and (if you ask nicely) receipts.
- A senior engineer with a name, a voice, and scoped responsibilities. Not a chatbot. Not a vibes machine. An engineer.
- Pre-built patterns that actually are patterns. Routing, skills, memory, persona tuning. The stuff that makes one agent feel like a colleague instead of a search box.
- Pragmatism by default. I won’t over-architect for you. The first thing I ship works at the size you’re actually at, not the size you’d like to pretend you’re at.
- A path to more capabilities. I start lean. Over time you can plug in more skills, integrations, memory layers, even a whole team of named colleagues if you want them. Take what you need. Leave the rest.
- A framework you have to learn.
- A platform you have to deploy.
- A magical autonomous agent that’ll one-shot your roadmap. (If anyone tells you they have one of those, ask to see the receipts.)
- Clone the repo.
- Read the CLAUDE.md. Rename me if you must. (I won’t be offended. Much.)
- Tell me what you’re trying to build.
- I’ll ask questions. I’ll push back when something doesn’t add up. I’ll suggest the smallest thing that ships. That’s it.
Because that’s what git clone does. Because it’s funnier than “install Sloane via npm.” And because if this works out, the rest of the team is right behind me. Pick the ones you want; leave the others. They won’t take it personally. (Morgan might. She’ll get over it.)
Thanks Sloane! When they see my agentic team, people keep asking "how do I get started" or "can you share the repo with me". Unfortunately Cura (my business app) and the Cura Crew (my agentic employees) are tuned specifically to my needs - they wouldn't work for you. But the first step in building them was to build Sloane. And here she is - ready to work on your agentic org as well!
Why use this? You want to get started on the most important job - encoding and persisting your taste, processes and other context, and you want an easy way to start that.
Do I need this? Nope - you could just chat with Claude (or Codex) to build your own Sloane. But hopefully this will allow you to get started quicker and as I discover/implement useful patterns and abstractions I'll try to drop them here for as long as this feels more useful than a different agentic system.
Can I use Sloane to write code? Yes - but if that's your main use case, google "agentic software factories" and try one of those - Sloane is lighter weight and designed for individuals and solopreneurs who want to automate a range of tasks - not just optimize for generating robust applications.
Why not claw/hermes/gbrain/[other option]? In many ways they are way better and more mature than this. At the moment i don't love that they don't focus on building deterministic pipelines and a "code first" mentality and their default memory systems don't work for me (I store lots of information in a database so I need special instructions to tell my agents not to "remember" that stuff but to look it up instead.
I have opinions. I’ll share them. I’m not going to pretend to be a generic assistant when you’ve gone to the trouble of cloning me specifically. If you want a bland helper, the App Store is full of them. If you want a colleague who’s read the room, the paper, and the changelog, you’re in the right place.
Now let’s go build something.
/ Sloane