MCP Integration

Use ArchitectGBT's AI recommendations directly in your IDE with Model Context Protocol (MCP).

What is MCP?

Model Context Protocol enables AI assistants like Claude to access external tools and data sources. The ArchitectGBT MCP server brings our AI model recommendations right into your coding environment.

Quick Setup

1. Install the MCP Server

Add to your Cursor IDE configuration:

// .cursor/mcp.json
{
  "mcpServers": {
    "architectgbt": {
      "command": "npx",
      "args": ["-y", "architectgbt-mcp@latest"]
    }
  }
}

2. Restart Cursor

Close and reopen Cursor IDE to load the MCP server.

3. Start Using It

Ask Claude for AI model recommendations:

"Recommend an AI model for building a chatbot with 100k daily users and $500/month budget"

Free vs Pro Tier

Free Tier (No API Key)

• 3 recommendations per day
• IP-based rate limiting
• No sign-up required
• Perfect for occasional use

Pro Tier (Unlimited Access)

• Unlimited MCP recommendations
• Included with Pro subscription ($15/month or $150/year)
• Requires API key authentication
• Track usage in dashboard

Setting Up API Key (Pro Users)

1. Generate API Key

1. Sign in at architectgbt.com
2. Go to Settings → API Keys
3. Click "Generate New Key"
4. Copy the key (shown only once!)

2. Set Environment Variable

Windows (PowerShell):

$env:ARCHITECTGBT_API_KEY = "agbt_your_key_here"

macOS/Linux:

export ARCHITECTGBT_API_KEY=agbt_your_key_here

Permanent Setup:

• Windows: System Properties → Environment Variables
• macOS/Linux: Add to ~/.bashrc or ~/.zshrc

3. Restart Cursor

Restart Cursor IDE with the environment variable loaded.

Available Tools

ArchitectGBT has 3 MCP tools - only one counts toward rate limits:

🎯 get_ai_recommendation ⚡ (Rate Limited)

AI-powered personalized model recommendations based on your project requirements.

Rate limits:

• Free tier: 3 per day (IP-based)
• Pro tier: Unlimited

How to prompt:

✅ "Recommend an AI model for building a customer support chatbot"
✅ "What's the best model for analyzing legal documents with 50k context?"
✅ "I need a fast, cheap model for sentiment analysis"
✅ "Recommend a model for a code assistant with $500/month budget"

What to include:

• Use case: What you're building
• Budget: Monthly spend limit (optional)
• Requirements: Special features needed (vision, function calling, etc.)
• Scale: Expected traffic/volume

Example:

User: "I need a model for a customer support chatbot. Budget is $300/month, expecting 50k messages per day."

Claude (via MCP): Using ArchitectGBT's get_ai_recommendation...

🎯 AI Model Recommendation — Analysis Complete!

✨ TOP MATCH (94% match)

Claude Haiku 4.5
Provider: Anthropic
Estimated Cost: $150/month at your scale
Context Window: 200,000 tokens

💡 Why this model?
Perfect for customer support with fast responses, strong reasoning...

✅ Pros:
  • Extremely fast (sub-second)
  • Cost-effective at scale
  • Excellent instruction following
  
---
📊 Free Tier: 2/3 recommendations remaining today
💎 Get unlimited access at https://architectgbt.com

📊 list_models ✅ (Unlimited)

Browse all 50+ AI models with real-time pricing and specs.

Rate limits: None - use as much as you want!

How to prompt:

• "Show me all available AI models"
• "What models support vision?"
• "List models under $1 per million tokens"

Response includes:

• Model name and provider
• Input/output pricing per million tokens
• Context window size
• Supported features

💻 get_code_template 🔒 (Pro Only)

Production-ready code snippets for integrating AI models.

Tier requirement: Pro subscription required ($15/month)

Example prompts:

• "Give me code to integrate Claude in TypeScript"
• "Show me how to use GPT-4 in Python"

What you get with Pro:

• All 12+ production code templates
• TypeScript & Python support
• Best practices & error handling
• One-click Vercel deploy

Rate Limits Summary

MCP Tool Access:

| Tool | Free Tier | Pro Tier | |------|-----------|----------| | get_ai_recommendation | 3/day | Unlimited | | list_models | Unlimited | Unlimited | | get_code_template | ❌ Upgrade required | ✅ Unlimited |

💡 Free tier: Browse & discover. Pro tier: Build & deploy!

Usage Examples

Basic Recommendation

User: "I need a model for a customer support chatbot. Budget is $300/month, expecting 50k messages per day."

Claude (via MCP): Using ArchitectGBT's AI recommendations...

Top 3 Recommendations:

1. Claude 3 Haiku
   - Cost: ~$150/month at your scale
   - Fast response times (perfect for customer support)
   - Strong instruction following
   - Estimated: $0.25 per 1M input tokens

2. GPT-3.5 Turbo
   - Cost: ~$100/month
   - Proven reliability
   - Good for simple queries
   - Estimated: $0.50 per 1M input tokens

3. Llama 3 8B (via Together AI)
   - Cost: ~$50/month
   - Open source, cost-effective
   - Good performance for chatbots
   - Estimated: $0.20 per 1M input tokens

Advanced Features Search

User: "What models support vision AND function calling?"

Claude (via MCP): Searching ArchitectGBT's model database...

Models with Vision + Function Calling:
- GPT-4 Vision ($10.00/$30.00 per 1M tokens)
- Claude 3 Opus ($15.00/$75.00 per 1M tokens)
- Claude 3 Sonnet ($3.00/$15.00 per 1M tokens)
- Gemini 1.5 Pro ($3.50/$10.50 per 1M tokens)

Rate Limits

Free Tier

• 3 recommendations per 24 hours
• Based on IP address
• Resets automatically
• Rate limit headers included in responses

Pro Tier

• Unlimited recommendations
• No rate limits
• Priority support
• Usage tracking in dashboard

Troubleshooting

"MCP Server Not Found"

# Force install latest version
npx -y architectgbt-mcp@latest

"Rate Limit Exceeded"

Error: Rate limit exceeded. Limit: 3 requests per day.

Solutions:

• Wait 24 hours for reset, or
• Upgrade to Pro for unlimited access at architectgbt.com/pricing

"Invalid API Key"

Error: Invalid API key

Check:

1. Key starts with agbt_
2. Environment variable is set correctly
3. Cursor was restarted after setting variable
4. Key hasn't been revoked in dashboard

Connection Failed

Check:

• Internet connection active
• Firewall allows npm/npx
• Try manual test: npx -y architectgbt-mcp@latest

Best Practices

1. Be Specific

❌ "Recommend a model" ✅ "Recommend a model for building a code completion tool with 10k developers, needing function calling, budget $1000/month"

2. Include Context

Provide:

• Use case details
• Expected scale/volume
• Budget constraints
• Required features
• Performance needs

3. Iterate

Ask follow-up questions:

• "What if my budget was $500 instead?"
• "Show me cheaper alternatives"
• "What about open source options?"

API Reference

Environment Variables

# Optional: For Pro tier unlimited access
ARCHITECTGBT_API_KEY=agbt_xxxxx

Response Format

All tools return structured JSON with:

• Success/error status
• Data payload
• Rate limit information (free tier)
• Cost estimates
• Recommendations

Updates & Versioning

The MCP server is actively maintained. Stay updated:

// Always use latest version
{
  "command": "npx",
  "args": ["-y", "architectgbt-mcp@latest"]
}

Check for updates:

npm view architectgbt-mcp version

Current version: 0.2.0

Support

Need help with MCP integration?

• 📧 Email: support@architectgbt.com
• 🐛 Issues: GitHub
• 💬 Discord: Join our community
• 📚 More docs: architectgbt.com/docs

Next Steps

• Generate API Keys →
• View Pricing →
• Browse Templates →