Nants
PT EN

Using Nants

From your first connection to your first app.

First connection

  1. Open Settings and add a connection: a name, the base URL and the API key for your provider. Any OpenAI compatible API works.
  2. Under Agents, pick the model for the Architect, the Developer and the Reviewer.

If something is missing on your machine, the Getting started window tells you what.

First project

  1. Create a project. It becomes a folder in ~/NantsProjects, with Git.
  2. In Conversation, describe what you want to build.
  3. Answer the Architect, read the Plan and approve it.
  4. Follow along on the Board. When it is done, open the App and try it.

To open Nants and the app on your phone, see Use it from your phone.

The windows

Screenshot of the Nants desktop with a finished project: the Final report on the left, the App window with the Aurora Cafe website running, and below it the Conversation with the Architect and the Run window with the server log.
A finished project: Report, App, Conversation and Run side by side.

Conversation

You and the Architect.

Plan

The tasks, reviewed before any work starts.

Board

Every task, from queue to delivery.

Activity

What each agent is doing right now.

App

Your app running, ready to try.

Costs

What each part cost.

Costs

The Costs window shows the project total and the cost per task, per agent and per model, as it happens. You can also set a budget.

Figures are estimates based on each model's price. Your provider does the actual billing.

Screenshot of the Nants Costs window: an estimated total of US$ 0.31, with images, Reviewer and Developer listed separately, and links to see the details per task and model and to set a budget.

Where your data lives

~/.config/nants
Settings and connections, including your API keys.
~/.local/share/nants
History, logs and costs.
~/NantsProjects
Your projects, each with its own Git repository.

Everything stays on your computer. Only the agents' requests go to the AI provider you chose.