Plain Files  ·  Your Repository  ·  Inspectable Output

Continue instead of reconstruct.

Anchor is an approach to AI-assisted engineering that helps projects stay coherent as they evolve across sessions, tools, and contributors.

The Problem

Projects become harder to understand long before they become harder to code.

Conversations are enough at first. Decisions pile up. A different tool or a new session suggests something incompatible with what was already settled. More time goes into restoring context than building.

When project understanding lives in chat history, every new session starts by reconstructing what the project already knew.

The Approach

Anchor organizes project understanding so humans and AI can build from the same foundation.

Like version control and CI/CD before it, Anchor makes an engineering constraint explicit — durable project context — so collaborators stop reconstructing it from scratch.

No magic

Anchor doesn’t become your documentation. It becomes the map to it.

How it works

What that looks like

Try it

Install, initialize, then connect your AI host.

npm install -g @jon4ohio/anchor-runtime

cd your-project
anchor init

`anchor init` creates `.anchor/config.json` — it declares what your project coordinates. It does not activate your coding agent.

Next: connect your AI host so it can invoke Anchor’s capabilities, verify with an orientation ask, then complete your first coordination (record an ADR).

A few principles

• Projects outlive conversations.
• Durable artifacts beat chat history.
• Explicit responsibilities reduce ambiguity.
• Continue instead of reconstruct.
• Map existing documentation — don't own it.