Get Started

Up and running in minutes.

Ada runs in your terminal. No account. No dashboard. Intent in, governed artifacts out.

Prerequisites

Node.js 18 or later — nodejs.org
pnpm — npm install -g pnpm
An Anthropic API key — console.anthropic.com

Install

bash

pnpm add -g @motherlabs/ada

Then set your API key:

bash

ada config set-key --env ANTHROPIC_API_KEY

Run Ada

In your project directory:

bash

ada compile

Ada opens an elicitation session. It will ask the minimum set of questions needed to resolve blocking unknowns. Answer at your level of abstraction — not technical, semantic. What should the system do? What should it not?

When elicitation is complete, Ada compiles the canonical semantic model and hands it to Claude Code. The governed blueprint is in .ada/.

What to expect

Ada streams its reasoning in real time — you can see it thinking, not just a spinner.

Elicitation asks 3–7 questions depending on intent complexity. No technical questions. Semantic ones.

Compilation produces artifacts in .ada/ — intent graph, entity map, process flow, blueprint.

Claude Code spawns with the governed blueprint. It builds against Ada's decisions, not against memory.

What's next

Read the Ada architecture on the Ada page to understand how the pipeline works. The source is on GitHub.