v0.1 · Loop kit for Next.js SaaS

Your users ask for a feature. An AI builds it. You merge.

From request to a working preview, on mock data until you say ship.

A user requests a word counter, answers two questions, confirms a spec, tries it on a private mock-data preview, asks for a fix and approves it
Recorded from examples/demo-notes. The Claude and GitHub calls are stubbed for the recording; the widget, server, preview link and mock-data preview are the real ones.

// the loop

  1. 01A user clicks Request a feature.
  2. 02An AI interviews them and writes a spec they confirm.
  3. 03A GitHub Action has Claude Code build it on a branch.
  4. 04They try it on a private preview with mock data, and ask for fixes.
  5. 05You merge. Only then does it run on real data, behind a flag that turns on for them first.

// safety model

Mock data until release. Preview deployments have no production env vars and no path to your database.

Policy in code, not prompts. Specs that touch auth, billing, migrations, secrets or CI are declined.

One human gate. Nothing reaches production data until you merge.

// license

totalmod is dual-licensed. Use it free under the AGPL-3.0, or get a commercial license for closed-source products. Code that npx totalmod init adds to your repo is MIT-0, so it's yours.