Human Repo

Your context, compounding — in every AI. → humanrepo.com

You have a repo for your code. You have nothing for yourself. So every AI you use — ChatGPT, Claude, Gemini — meets you as a stranger, every time. You re-explain your stack, your projects, who you are. Nothing compounds. Their "memory" is a siloed gray blob you don't own and can't move.

Human Repo is the fix: a portable, version-controlled repository of you — what you're doing, what you want, how you work — that you own and carry into any AI, so you never start from zero again.

Why it's different

• You own it. Plain markdown in a git repo. Portable, inspectable, forkable, deletable. Not trapped in anyone's product.
• It compounds. Every session can propose a patch; you merge. It gets truer over time instead of staler.
• It's yours to author. Nothing becomes "true about you" without your signature. (See the principle in ../docs/the-mission.md.)
• It works everywhere. Compile it into portable context for any AI today; native injectors next.

How it works (corpus + tools + loop)

1. Your repo — a folder of markdown you own. (FORMAT.md, template/)
2. Compile — humanrepo build turns it into a context block you paste into any AI (or save as custom instructions / project knowledge).
3. Compound — after a session, merge what's new. It grows with you instead of resetting.

Status

v0.0.1 — the open format, a zero-dependency compiler, and an MCP server so Claude loads your repo automatically (mcp/). Dogfooding now. Next: a browser injector (ChatGPT / Gemini web), and humanrepo merge to apply reviewed proposals from INBOX.md.

Try it

node bin/humanrepo.mjs build template
cat template/dist/context.md

The product is Human Repo. The principle we never violate is that you keep the wheel — see ../docs/the-mission.md.