---
title: "Choosing an AI provider"
description: "HolaClaw isn't tied to a single AI model. You can power your assistant with Claude, GPT, Gemini, Grok, DeepSeek, OpenRouter, or a model running locally on your Mac. This guide helps you pick."
updated: 2026-06-11
canonical: https://holaclaw.ai/docs/ai-providers/choosing-a-provider
---

When you create an assistant in HolaClaw, one of the first decisions is which AI model powers it. Each provider has its own personality, pricing, and trade-offs — and the right answer depends on what you're building.

This page is the short version: a side-by-side comparison and a one-paragraph feel for each provider. Every provider also has its own guide with step-by-step setup, model recommendations, and pricing details — follow the links below when you have narrowed it down.

> **Tip — the short version.** If you're just starting out, pick **Anthropic Claude**. It's a great all-rounder, handles long conversations well, and works beautifully with HolaClaw's memory system. You can always use a different provider for your next assistant.

![The "Connect the brain behind your assistant" step of the Create Assistant flow, showing the provider list — Local, GPT, Claude, Gemini, Grok, DeepSeek, OpenRouter — in the left sidebar.](/images/docs/ai-providers/ai-providers.webp)

## Quick comparison

| Provider | Auth | Best for | Cost | Runs locally? |
| --- | --- | --- | --- | --- |
| **[Claude](/docs/ai-providers/anthropic-claude)** | API key | Long conversations, nuance, writing | $$ | No |
| **[OpenAI](/docs/ai-providers/openai)** | ChatGPT subscription **or** API key | General use, broad knowledge | $$ | No |
| **[Gemini](/docs/ai-providers/google-gemini)** | API key | Multimodal, long context, low cost | $ | No |
| **[Grok](/docs/ai-providers/grok)** | API key | Snappy answers, distinct voice | $$ | No |
| **[DeepSeek](/docs/ai-providers/deepseek)** | API key | Capable chat at very low cost | ¢ | No |
| **[OpenRouter](/docs/ai-providers/openrouter)** | API key | Trying many models with one key | varies | No |
| **[Local](/docs/ai-providers/local-models)** | None | Privacy, offline use, no API costs | Free | Yes |

## Anthropic Claude

Claude is our default recommendation. It has a warm, conversational feel that pairs well with HolaClaw's memory system — it picks up on small details you mention and remembers them naturally across sessions. Great for journaling companions, writing partners, and travel planners (like Luna). Authentication is by API key.

**[Set up Claude →](/docs/ai-providers/anthropic-claude)**

## OpenAI

The most familiar option, and the only provider with two ways in: sign in with the **ChatGPT subscription** you already pay for (Plus, Pro, or Team), or bring a platform **API key** and pay per token. GPT models are versatile, fast, and excellent all-rounders.

**[Set up OpenAI →](/docs/ai-providers/openai)**

## Google Gemini

The budget-friendly option with a huge context window. Pick Gemini if you want low API costs, plan to send long documents or many images, or want an assistant that can hold a lot of context at once. Authentication is by API key from Google AI Studio.

**[Set up Gemini →](/docs/ai-providers/google-gemini)**

## Grok (xAI)

Grok has a distinct, looser voice and tends to answer with confidence on current events. Pick it if you want an assistant with a different personality from the default cloud models. Authentication is by API key.

**[Set up Grok →](/docs/ai-providers/grok)**

## DeepSeek

Remarkably capable models for the price — a fraction of the equivalent OpenAI or Anthropic tier. Pick DeepSeek if cost matters most and you're willing to trade a little polish for very cheap tokens. Authentication is by API key.

**[Set up DeepSeek →](/docs/ai-providers/deepseek)**

## OpenRouter

Not a model maker but a router: one API key that unlocks models from Anthropic, OpenAI, Google, xAI, DeepSeek, and more. Pick OpenRouter if you want to experiment across providers without juggling half a dozen accounts and keys.

**[Set up OpenRouter →](/docs/ai-providers/openrouter)**

## Local models

If privacy matters most, run a model entirely on your Mac. No data ever leaves your machine — no API keys, no cloud, no usage bills. HolaClaw ships with an embedded runtime, so there's nothing extra to install; models download on demand from inside the app. The trade-off is speed and quality, and you'll want an M-series Mac with 16GB+ of unified memory.

**[Set up a local model →](/docs/ai-providers/local-models)**

## Not locked in

You choose the provider and model whenever you create an assistant, and each assistant can use a different one — a local model for a private journal, Claude for a writing partner.

One thing to know up front: **the choice is made once, at creation**. There's currently no way to change the provider or model of an existing assistant. If you change your mind, create a new assistant with the provider you want — and since assistants are quick to set up, that's less of a chore than it sounds.

Still not sure? [Ask on Discord](https://discord.gg/FbxAbS5sGQ) — the community is happy to share what they're using for what.