Skip to content

AI profiles

An AI profile is a saved model configuration — a provider, an endpoint, and a key — that you can switch between at will. Because Altretta ships no LLM of its own, profiles are how you bring the model you choose: a frontier cloud model for hard questions, a local model for privacy, or several, side by side.

The AI is yours, on your subscription. Altretta never sits between you and your provider taking a cut of tokens — it just points the grounded, cited query flow at whichever model you've selected.

Presets for the models you use

Altretta includes ready-made presets so you can add a model in seconds:

Anthropic (Claude)

Connect your Anthropic account and its Claude models.

OpenAI

Connect an OpenAI account and its models.

OpenRouter

Reach many providers through a single OpenRouter key.

Gemini

Use Google's Gemini models.

DeepSeek

Connect DeepSeek's models.

xAI (Grok)

Connect xAI's Grok models.

Groq

Fast inference through Groq.

Mistral

Connect Mistral's models.

Z.AI

Connect Z.AI's models — with a separate preset for the Z.AI Coding Plan endpoint.

Ollama (local)

Run a model entirely on your own machine — nothing leaves your computer.

Custom

Any OpenAI-compatible endpoint: give it a base URL, a model name and a key.

The preset is a convenience, not a constraint — what actually matters is the base URL, the model name and the key. Any provider that speaks the OpenAI chat-completions shape works through the Custom preset.

Want maximum privacy? Pair a local Ollama profile with your local vault and your notes never leave your device — grounded answers, no cloud round-trip.

Save, rotate, and toggle

Profiles are made to be plural. Keep as many as you like and move between them without re-entering keys.

1

Add a profile

Pick a preset (or configure a custom endpoint), name it, and paste your key. Altretta tests the profile when you save it, so you find out immediately if the key or endpoint is wrong. The test is non-blocking: the profile is saved either way, and you can fix it later.

2

Rotate between models

Switch the active profile whenever you want — a strong reasoning model for a thorny question, a fast or local one for quick lookups.

3

Enable or disable per profile

Turn individual profiles on or off. Keep a profile saved but inactive without deleting it.

Profiles and grounding work together

Whichever profile is active, the query-first, grounded protocol is the same: the model queries your vault, answers from cited passages, and is held to the groundedness verdict. Swapping models changes who answers — it never changes the rule that answers must be grounded and cited. See Grounded, cited answers.

Profiles power the in-app Ask panel and Chat. To let an external client like Claude Desktop or Cursor answer from your vault instead, connect it over MCP — see Connect your AI.

Next step

Chosen your model? → Grounded, cited answers to start asking, or What your AI sees to review the context you're exposing.