Make AI Model Manager

A tiny local web app to scan and bulk-update AI model versions (Claude, Gemini) across all your Make.com scenarios. No more opening 47 scenarios one-by-one to change claude-3-5-sonnet to claude-sonnet-4-6.

• Zero dependencies — pure Node.js (uses only stdlib).
• One JS file + one JSON config. Move the folder, it moves with you.
• Uses the official Make.com API — no browser automation, no scraping.

What it does

1. Connects to your Make.com account with your API token.
2. Scans every scenario in the selected organization. For each AI module it finds, it checks whether the model ID matches one of your rules.
3. Shows you exactly which scenarios + modules need an update.
4. Patches them in bulk, in one click. Full blueprint PATCH — all your other settings stay intact.

Supported rule shape

Each rule has one or more from patterns (regex, case-insensitive) and a single to target. Example:

{
  "from": ["^claude-opus"],
  "to": "claude-opus-4-7",
  "toLabel": "Claude Opus 4.7"
}

This sweeps any model starting with claude-opus (e.g. claude-opus-4-1-20250805, claude-opus-4-5-20251101) to claude-opus-4-7.

Default rules ship for Claude Opus / Sonnet / Haiku and Gemini Pro / Flash. Add your own via the UI.

Setup

Prereq: Node.js 18+.

git clone https://github.com/arielmoatti/make-ai-model-manager.git
cd make-ai-model-manager
npm start

Open http://localhost:3000. You'll be prompted to paste your Make.com API token — create one at https://<your-zone>.make.com/users/me/api (needs scenarios:read and scenarios:write). It's stored locally in make-ai-model-manager.secret.json next to the script and never leaves your machine.

Custom port

node make-ai-model-manager.js --port=8080

Files

File	Purpose	Gitignored
make-ai-model-manager.js	The entire app (server + embedded HTML + client JS)	no
make-ai-model-manager.json	Your rules + last-used org/team	no (ships with defaults)
make-ai-model-manager.secret.json	Your API token	yes

Safety notes

• The app validates the token against Make.com on save — bad tokens are rejected before they're stored.
• Every PATCH goes to /api/v2/scenarios/{id} with the full blueprint. Make.com refuses the PATCH if the blueprint references deleted connections or webhooks — so broken scenarios are surfaced, not silently corrupted.
• Rate limits are respected: exponential backoff (0 / 1.5s / 4s / 10s / 25s), honors Retry-After headers, paces 400ms between scenarios.
• Pre-existing broken scenarios (dangling connection/hook references) will fail with a clear error — not caused by this tool.

Screenshots

(add when published)

License

MIT — see LICENSE.