Introducing ContextStore: a native Mac app for context repositories

Most of the way people work with AI today is through prompts. But there's only so much you can cram into a prompt. Simple prompts lead to sub-par results that require lots of iteration to get right. Not because AIs aren't intelligent, but because you haven't shared the full context of what you're working on with them.

Just like you wouldn't onboard a new employee without giving them access to critical documents — product requirements, marketing guidelines, customer profiles — you need to give AIs the right context to make good decisions. I wrote about this idea last week and called it a context repository.

The problem is that documents are often locked away in places where AIs can't get to them. Google Docs, Notion, Confluence. You can work around this with connectors and MCPs and all kinds of technical solutions, but it turns out the best way to do this is with a folder full of Markdown files. Markdown is plain text, making it easy for AIs to read and modify. It consumes fewer tokens. And if you want to truly collaborate with AIs on documents, it's the best format.

But not everyone knows Markdown. And there aren't easy ways to share it.

I started working on a Mac app for this back in January. I've been building nights and weekends ever since, fitting it in around my full-time work at Gierd. Today, I'm announcing ContextStore — a native Mac app for creating and managing context repositories. It's in beta and I'm looking for testers.

ContextStore gives you a Markdown editor, documents organized into Spaces and Projects, and automatic GitHub sync. No terminal or fancy technical tools needed. You write, and your context repository stays current. One app, designed so anyone on your team — technical or non-technical — can create and edit content.

Three principles drove the design. First, it's local-first. Your documents live on your machine as plain Markdown files. No cloud service between you and your data. Second, it's GitHub-native. Connect a Space to a repo and ContextStore commits and pushes automatically. Merge conflicts get a visual resolution panel — no rebasing required. Third, there's no lock-in. It's Markdown all the way down. Your context repository works with Claude Code, Cursor, Copilot, ChatGPT — anything that reads files.

I've been dogfooding ContextStore to manage the planning documents for ContextStore itself — vision, roadmap, customer profiles, competitive analysis. When I sit down with Claude Code to work on a feature, the agent reads the context repository and already knows what I'm building and why. No preamble. It just works.

If you're looking for a better way of working with AI, I'd love for you to try the beta.

Fill in your details below to join the Beta! 👇