Choosing an AI Provider

An "AI provider" is simply the company (or software) that runs the AI model you're chatting with. Think of it like choosing a phone carrier — you pick the one that best fits your needs.

ClawDesk supports 10+ AI providers, and you can use several at the same time.

---

Quick Recommendation

Not sure where to start? Here's a simple decision tree:

---

Provider Comparison

Here's a plain-English comparison of the main AI providers:

Provider	Best For	Cost	Speed	Quality	Privacy
Ollama	Free local AI	Free	Fast	Good	Excellent (local)
ClawDesk Local	Offline AI	Free	Fast	Good	Excellent (local)
Claude (Anthropic)	Writing, analysis, coding	Pay per use	Fast	Excellent	Good
ChatGPT (OpenAI)	General purpose	Pay per use	Fast	Excellent	Good
Gemini (Google)	Long documents, multimodal	Pay per use	Fast	Very Good	Good
Azure OpenAI	Enterprise/business	Pay per use	Fast	Excellent	Very Good
AWS Bedrock	Enterprise/AWS users	Pay per use	Fast	Excellent	Very Good
Cohere	Search, enterprise AI	Pay per use	Fast	Good	Good
OpenRouter	Multi-model access	Pay per use	Varies	Varies	Good
Vertex AI	Google Cloud users	Pay per use	Fast	Very Good	Very Good

---

Option 1: Free & Local (No API Keys Needed)

Ollama — The Easy Local Option

Ollama lets you run AI models on your computer for free. It's the easiest way to get started without paying anything.

Pros:

• Completely free
• Runs on your computer (private)
• Works offline after initial download
• Easy to set up

Cons:

• Needs a decent computer (8GB+ RAM recommended)
• Quality depends on the model you choose
• Slower than cloud providers on older hardware

How to set up Ollama:

1. Download Ollama from ollama.com
2. Install it (just like any other app)
3. Download a model (open your terminal and type):

   ollama pull llama3.1

4. In ClawDesk, go to Settings → Providers → Ollama
5. Base URL: http://localhost:11434 (leave as default)
6. Default Model: llama3.1
7. Click Save

That's it! You can now chat with an AI for free, running entirely on your computer.

Recommended Ollama Models

Model	Size	Good For	RAM Needed
llama3.1	4.7 GB	General chat, writing	8 GB
llama3.1:70b	40 GB	Best quality (local)	64 GB
codellama	3.8 GB	Programming help	8 GB
mistral	4.1 GB	Fast general chat	8 GB
gemma2	5.4 GB	Google's model	8 GB

ClawDesk Local Models — Built Right In

ClawDesk can also run AI models directly, without Ollama or any external software.

1. Go to the Local Models page in ClawDesk
2. ClawDesk will detect your hardware
3. It recommends models that will run well on your machine
4. Click Download on a model
5. Once downloaded, click Start
6. The model appears as a provider option in your chats

---

Option 2: Cloud Providers (Best Quality)

Cloud providers run their AI models on powerful servers, giving you the best quality results. You'll need to create an account and get an "API key" (like a password for the service).

What's an API Key?

An API key is a unique code that identifies you to the AI service. It looks something like this:

sk-ant-api03-xyzABC123...

How to keep it safe:

• Never share your API key publicly
• Don't post it on the internet
• Store it in ClawDesk's settings (it's encrypted)

---

Anthropic (Claude)

Claude is known for being helpful, harmless, and honest. It's excellent at writing, coding, and analysis.

Getting started:

1. Go to console.anthropic.com
2. Create an account
3. Add a payment method (you pay only for what you use)
4. Go to API Keys → Create Key
5. Copy the key
6. In ClawDesk: Settings → Providers → Anthropic → Paste your key
7. Default model: claude-sonnet-4-20250514

Pricing: ~$3-15 per million tokens (roughly $0.01-0.05 per long conversation)

Best for: Writing, analysis, coding, following instructions carefully

---

OpenAI (ChatGPT / GPT-4)

The most well-known AI provider. Powers ChatGPT.

Getting started:

1. Go to platform.openai.com
2. Create an account
3. Add a payment method
4. Go to API Keys → Create new key
5. Copy the key
6. In ClawDesk: Settings → Providers → OpenAI → Paste your key
7. Default model: gpt-4o

Pricing: ~$2.50-10 per million tokens

Best for: General conversations, data analysis, creative writing

---

Google Gemini

Google's AI model, good at processing long documents and multimodal content (text + images).

Getting started:

1. Go to aistudio.google.com
2. Click "Get API Key"
3. Create or select a project
4. Copy the API key
5. In ClawDesk: Settings → Providers → Google → Paste your key
6. Default model: gemini-2.0-flash

Pricing: Free tier available, then ~$0.075-1 per million tokens

Best for: Long documents, image analysis, integration with Google services

---

Option 3: Enterprise Providers

These are for businesses and organizations that need extra security, compliance, and support.

Azure OpenAI

• Same GPT models, hosted on Microsoft Azure
• Better data privacy controls
• Enterprise SLAs

AWS Bedrock

• Access to multiple AI models through AWS
• Use your existing AWS account
• VPC security and compliance

Vertex AI (Google Cloud)

• Google's enterprise AI platform
• Advanced customization options
• Google Cloud integration

---

Using Multiple Providers

One of ClawDesk's best features is that you can use multiple providers at the same time:

Why use multiple providers?

• Use the free local model for everyday chatting
• Switch to Claude for important work that needs high quality
• Use Gemini when you need to process very long documents
• Compare answers from different AIs to get the best results

How to switch between providers:

1. In any chat, look at the model dropdown at the top
2. Click it and select a different model
3. Your conversation continues with the new model

---

Smart Routing & Failover

ClawDesk includes smart failover — if one AI provider is down or gives an error, ClawDesk automatically tries another:

This means you never get stuck if one service goes down. ClawDesk handles it automatically.

---

How Much Does It Cost?

Understanding Tokens

AI providers charge based on "tokens" — small pieces of text. As a rough guide:

• 1,000 tokens ≈ 750 words (about 1.5 pages of text)
• A typical conversation uses 1,000-5,000 tokens
• That costs roughly $0.01-0.05 per conversation

Monthly Cost Estimates

Usage Level	Description	Estimated Cost
Light	~10 conversations/day	$5-15/month
Moderate	~30 conversations/day	$15-40/month
Heavy	~100+ conversations/day	$40-100/month
Local only	Unlimited (Ollama)	Free

Cost Saving Tips

1. Use Ollama for casual conversations (free)
2. Only use cloud providers for important tasks
3. Choose smaller/faster models when speed matters more than quality
4. ClawDesk's smart routing helps avoid wasting tokens on errors

---

Provider Capabilities

Not all providers support all features:

Feature	Ollama	Claude	GPT-4	Gemini
Text chat	✅	✅	✅	✅
Streaming	✅	✅	✅	✅
Tool use	✅	✅	✅	✅
Vision (images)	Some models	✅	✅	✅
Code execution	❌	✅	✅	✅
JSON mode	✅	✅	✅	✅
Long context	Varies	200K	128K	1M+
Extended thinking	❌	✅	✅	✅
Cost	Free	$$	$$	$

---

Next Steps

• Running AI Locally → — Detailed guide to local AI setup
• Understanding Agents → — Create specialized AI assistants
• Privacy & Security → — How your data is protected