Skip to main content
Connect Your AI

Which AI connection should I use?

Compare API keys, subscription plans and local models to choose your first AI connection.

Written By Sujeevan Nagarajah

Last updated 22 days ago

Vinaa is the app you chat in; an AI connection provides the model that writes replies. You can add more than one connection and choose a different model for each chat.

Compare your options

An API key is a secret code that lets Vinaa use an AI provider through your account. A local model runs on your Mac rather than on a provider's computers.

Connection

Cost

Setup effort

Where messages are processed

Offline replies

API key: OpenAI, Anthropic or Google

Usually pay per use; credits or a limited free tier may be available

Create a provider account and paste a key

At the provider

No

ChatGPT Plan

Uses your plan's Codex allowance

Install Codex CLI, sign in and grant folder access

At OpenAI

No

Claude Plan

Uses your Claude plan; extra usage may cost more

Install Claude Code and sign in; Experimental

At Anthropic

No

Ollama or LM Studio, running a downloaded model on this Mac

No provider charge for local replies; uses your Mac's memory, storage and power

Install a separate app, download a model and run its server

On your Mac

Yes, after setup

Apple Foundation Models

No API key or per-message provider charge

Requires macOS 27 or later and Apple's on-device AI to be ready

On your Mac

Yes, once ready

OpenRouter

Model-dependent usage charges and credits

Sign in or paste a key, then choose models

Through OpenRouter and the selected model host

No

OpenAI-Compatible API

Depends on the service or server

Obtain its address, authentication details and models

Wherever that server runs

Only if the server and model are local

Local operation depends on using a downloaded model, rather than a cloud model. See the Ollama FAQ and LM Studio's offline guide. OpenRouter forwards requests to model hosts; review its privacy and billing information.

Choose a starting point

  • For a simple setup without Terminal: use an API key. Google's Gemini API has a free tier for eligible projects and models, with limits; check Google's pricing.

  • If you already have ChatGPT: consider ChatGPT Plan if you're comfortable following a few Terminal steps. Check OpenAI's current plan access and limits.

  • If you already have Claude: Claude Plan is Experimental and may stop working. For API access instead, choose Anthropic and manage API billing separately.

  • For offline work or local processing: choose a downloaded model in Ollama or LM Studio, or Apple Foundation Models on a supported Mac. Model size and your Mac's memory affect local performance.

  • To try models from many providers through one account: consider OpenRouter. Use OpenAI-Compatible API when a service gives you an OpenAI-compatible server address.

Cloud connections send the content needed for your request to the service. A local model keeps model processing on your Mac, but online tools still contact their services, and iCloud sync is a separate setting.

Find the connection choices

  1. Open Vinaa → Settings… (⌘,).

  2. Select AI Models.

  3. Click Add Connection in the toolbar.

  4. Choose a provider to begin its setup.

Note: Your Vinaa purchase or trial covers Vinaa. Any AI subscription, API credits or hosted-model charges are managed separately with the provider.

Related articles

  • Connect an AI provider

  • Get a Google Gemini API key

  • Use your ChatGPT plan in Vinaa

  • Run models locally with LM Studio