Product context for AI coding agents

The category

OriginAI is product context for AI coding. The slogan names the same thing: your missing product context layer.

Product context is the decided description of what to build — screens, states, roles, rules, and edge cases. A layer is where it sits: above the coding agent, below raw intent. Agents implement a published version of that layer. They do not reconstruct it from last week's chat.

This is not a PRD tool (those go stale), a design file (those skip states), or a coding agent (OriginAI feeds Claude Code, Cursor, and Codex). It is also not an LLM "context window" — that is tokens. Product context is a durable, versioned spec.

The idea

In OriginAI, product context is structured so people and coding agents can both read it and act on it. You refine it in the workbench and publish it as a release — a frozen snapshot. That snapshot is what agents load.

Why "context" instead of "docs"?

Traditional documents drift out of date, and chat disappears. Product context in OriginAI is different in three ways:

1. It's editable. You refine it directly in the workbench.
2. It's freezable. You publish it as a release that agents implement.
3. It's queryable. Agents can read it through the CLI — listing documents, fetching diffs, searching content, and so on.

What good product context includes

Good product context usually covers:

• The routes and screens, and the states each one can be in.
• Permissions and how the experience varies by role.
• The empty, loading, and error paths — not just the main flow.
• The business rules that code must follow and must not invent.
• What to build first, typically planned in the README.

Related

• What is OriginAI?
• Workspace vs release
• Review change requests
• Files and README