Before You Begin

You need almost nothing to start. The protocol is entirely plain text and files, and dotHumanize is tool-agnostic. Here is what you need before you begin.

Prerequisites

Each item below is a core requirement for bringing the protocol to life on a project — plus one optional convenience at the end.

AI Agent Environment

• An AI agent running in your working environment with full read/write access to your codebase. This is the only hard requirement: an agent you can point at your workspace folder that is permitted to read and modify files. (If your environment locks down local file access, you must clear those permissions first.)
• Note on continuity: Examples throughout this handbook standardize on Claude Code in an IDE for reference and clarity, but it is not a requirement. Any capable AI model that interacts with your local filesystem will work.

Target Workspace Folder

• A local folder to ground the protocol on. This can be an established codebase, a documentation or asset folder, a parent directory containing multiple sub-projects, or an entirely blank greenfield repository. dotHumanize adapts instantly to whatever folder composition it encounters.

Node.js Runtime

• A standard Node.js installation (the runtime that executes JavaScript files). dotHumanize uses a tiny, zero-dependency script to crawl your folder layout and collect deterministic facts.
• How to check: Run node --version in your terminal. If it prints a version number, you are set. It works identically across macOS, Windows, and Linux.
• Don't have it? The easiest path is to simply ask your AI agent to install it for you during your run. Alternatively, download the stable LTS build yourself from nodejs.org.

Spec-Driven Methodology

• A workflow built around explicit intent. dotHuman operates goal-by-goal rather than task-by-task. You state a goal, your agent drafts a short sequence (spec → plan → tasks), you review and adjust the plan, and then the agent executes.
• Note on tools: We reference SpecKit to anchor our structural examples, but you can use any framework or ticketing approach you already prefer (custom markdown templates, RFCs, Shape Up, or Jira issues).

Local and Private

• It all stays on your own machine. There are no accounts to create, no software to install beyond your AI tool, and nothing that phones home — the protocol is just plain files in a folder you control.

Code Editor or IDE (optional)

• Optional, but handy. An editor makes reviewing the agent's diffs and the growing .human/ trail easier to follow. Our examples use VS Code (with the agent in its integrated terminal), but any editor works — or none at all, if your agent runs entirely from the command line.

Next Steps

Once you have checked off the requirements above, you are ready to choose your entry path:

• For the five-minute overview, jump straight to the Quickstart.
• To see the foundational elements spin up, read Your First dotHuman.
• If you want a high-level conceptual primer first, start at the Introduction.

Before moving forward, please review the Disclaimer regarding independent protocols and tool naming.