Choosing an AI Provider

AI Companion supports four leading AI providers. Each has unique strengths, pricing, and characteristics. Choose the one that best fits your needs.

Supported Providers

OpenAI (GPT-4o)

Best for: Most users, reliable performance, industry standard

Model: GPT-4o
Strengths: Excellent quality, reliable performance, well-documented
Pricing: Pay-per-use, typically $0.01-0.03 per suggestion
API Keys: platform.openai.com/api-keys
Documentation: platform.openai.com/docs

OpenAI is the most widely used provider and offers consistent, high-quality responses. It's a great default choice if you're unsure which provider to use.

Anthropic (Claude Haiku 4.5)

Best for: Fast responses, cost efficiency, high-quality output

Model: Claude Haiku 4.5
Strengths: Fast response times, cost-effective, excellent quality
Pricing: Pay-per-use, typically $0.005-0.015 per suggestion
API Keys: console.anthropic.com
Documentation: docs.anthropic.com

Claude Haiku offers excellent performance at a lower cost than GPT-4o, making it ideal for high-volume support teams.

Grok (Grok-3)

Best for: Alternative option, xAI ecosystem users

Model: Grok-3
Strengths: Modern architecture, competitive performance
Pricing: Pay-per-use, pricing varies
API Keys: console.x.ai
Documentation: docs.x.ai

Grok is a newer option from xAI, offering competitive performance and modern features.

Gemini (2.5 Flash)

Best for: Google ecosystem users, cost-conscious teams

Model: Gemini 2.5 Flash
Strengths: Fast, cost-effective, Google integration
Pricing: Pay-per-use, typically very competitive
API Keys: aistudio.google.com/app/apikey
Documentation: ai.google.dev

Gemini offers fast, cost-effective AI responses with integration into Google's ecosystem.

Comparison Table

Provider	Model	Speed	Cost	Best For
OpenAI	GPT-4o	Fast	Medium	Most users, reliability
Anthropic	Claude Haiku 4.5	Very Fast	Low	High volume, cost efficiency
Grok	Grok-3	Fast	Medium	Alternative option
Gemini	2.5 Flash	Very Fast	Low	Cost-conscious teams

Pricing Considerations

How Pricing Works

You pay your provider directly - SupportRetriever doesn't charge for AI usage
Pay-per-use model - You're charged per API call (suggestion generation)
Typical costs - Most suggestions cost $0.005-0.03 depending on provider and message length
Monitor usage - Set up usage alerts in your provider account to avoid surprises

Estimating Costs

For a team receiving 100 customer messages per day:

Anthropic/Gemini: ~~$0.50-1.50 per day (~~$15-45/month)
OpenAI: ~~$1-3 per day (~~$30-90/month)
Grok: Varies, typically similar to OpenAI

Actual costs depend on:

Message length and complexity
Knowledge base usage (web search adds cost)
Provider pricing changes

Getting API Keys

OpenAI

Visit platform.openai.com/api-keys
Sign in or create an account
Click "Create new secret key"
Copy the key (starts with sk-proj- or sk-)
Set up billing and usage limits

Anthropic

Visit console.anthropic.com
Sign in or create an account
Navigate to API Keys section
Create a new key (starts with sk-ant-)
Configure billing and usage alerts

Grok

Visit console.x.ai
Sign in or create an account
Navigate to API Keys
Generate a new key (starts with grok-)
Set up billing information

Gemini

Visit aistudio.google.com/app/apikey
Sign in with Google account
Create API key (starts with AIza)
Configure usage limits and restrictions

Switching Providers

You can switch providers at any time:

Go to AI Companion settings
Click edit on the Provider field
Select a new provider
Enter the new provider's API key
Save changes

Your existing suggestions won't be affected, but new suggestions will use the new provider.

Provider-Specific Features

Knowledge Base Integration

All providers support knowledge base integration, but implementation varies:

OpenAI/Grok: Uses web_search tool with domain filtering
Anthropic: Uses web_search_20250305 tool with domain filtering
Gemini: Uses url_context tool

All methods work effectively, but you may notice slight differences in how knowledge base content is accessed and referenced.

Response Quality

All providers generate high-quality suggestions, but you may notice:

OpenAI: More detailed, comprehensive responses
Anthropic: Faster, more concise responses
Grok: Balanced between speed and detail
Gemini: Fast, practical responses

Recommendations

For New Users

Start with OpenAI or Anthropic:

Well-documented and reliable
Large user communities
Consistent performance

For High-Volume Teams

Consider Anthropic or Gemini:

Lower cost per suggestion
Fast response times
Good for scaling

For Cost-Conscious Teams

Try Gemini or Anthropic:

Most cost-effective options
Still excellent quality
Good performance

For Maximum Quality

Use OpenAI:

Industry-leading quality
Most comprehensive responses
Best for complex queries

Best Practices

Start with one provider - Test with your actual customer messages
Monitor costs - Set up usage alerts in your provider account
Compare quality - Try different providers if needed
Rotate keys annually - For security best practices
Use knowledge base - Improves accuracy regardless of provider