Introduction
dotHuman is an open, tool-agnostic operating layer that gives an AI agent a durable place to inherit how you work, survive context resets, and compound lessons over time.
What it is
dotHuman is a lightweight, plain-text operating layer that lives inside your project as a .human/ folder. It holds your project's working method — its goals, the understanding a fresh agent needs, the durable how-tos, and the running log. Written down once as Markdown, any AI agent you point at the project inherits how you work instead of re-figuring it out every session.
Because it's just Markdown and folder conventions, there's no framework and no lock-in — it drops into any workspace and works the same across any file-capable AI tool.
Why it exists
AI agents are transient. A fresh session starts with no memory of the last one, and hard-won context evaporates at every reset. The .human/ workspace is the durable place that survives those resets: a fresh agent re-grounds itself from it, inherits how you work, and accumulates lessons that compound instead of relearning the same things each time.
The goal isn't to take the human out of the loop — it's the opposite. The method keeps you steering the decisions, the verification, and the craft while the agent does more of the manual work. Be the dotHuman in the loop.
Naming
Three names are easy to mix up — keep them straight:
- dotHuman — the brand: the project, and what you're reading now (dothuman.org).
- .human — the method: the open, tool-agnostic protocol you teach an AI.
- .human/ — the workspace: the on-disk folder, dropped into your project root, where the method actually runs (you can nest one per scope).
Their own page goes deeper: Naming.
Where the name came from
Open a fresh repo and every tool's settings are tucked into semi-hidden dotfiles and dot-folders at the root — .env, .gitignore, .vscode, .github — each configuring a machine. Now that AI agents work in the codebase alongside us, the human needs one of their own. The .human/ folder is that anchor: the person's operational config, sitting right beside the machines'.
Start here
New here? Read these two first, in order:
- The Lifecycle — the loop you run on every goal.
- The Two Layers — your durable, project-wide rules, and the flexible per-goal layer where the work actually happens.
After those two, the rest is reference — browse it by theme in the sidebar, or look a term up A–Z in the Glossary.
Related
Last updated