architectgbt-mcp
v0.4.2
Published
Model Context Protocol server for ArchitectGBT - AI model recommendations with zero-friction onboarding
Maintainers
Readme
ArchitectGBT MCP Server
AI model recommendations directly in your IDE. Compare 50+ models, get code templates, and optimize costs without leaving Cursor or Claude Desktop.
Try it Free → · Get API Key → · View All Models →
Resources: NPM Package · GitHub Repository
Quick Start
No signup, no API key, no configuration needed:
npx -y architectgbt-mcpGet 3 free AI recommendations per day to try it out.
Want unlimited access + code templates + advanced features? Upgrade to Pro → (starts at $15/mo)
What is ArchitectGBT?
Choosing the right AI model shouldn't take hours of research. ArchitectGBT analyzes your project requirements and recommends the best model with actual cost estimates, not marketing fluff.
Instead of manually comparing pricing tables across OpenAI, Anthropic, Google, Meta, and Mistral, you get:
- AI-powered recommendations based on your project requirements
- Cost optimization with realistic usage estimates
- Model comparison across 50+ models with real-time pricing
- Production-ready code in TypeScript and Python
- Direct IDE integration via Model Context Protocol (MCP)
This MCP server brings ArchitectGBT's intelligence directly into your IDE, so you never have to leave your coding flow.
New to ArchitectGBT? Explore the full platform →
Why This vs Cursor's Auto?
Cursor's Auto feature is great for quick model switching, but it doesn't solve the research problem:
ArchitectGBT solves:
- Cost planning - Get realistic estimates before you build, not surprise bills after
- Model discovery - Find models you didn't know existed (Mistral, Cohere, specialized models)
- Requirements matching - Match your specific use case (context window needs, latency requirements, budget constraints)
- Production templates - Get working integration code, not just model names
When to use Cursor Auto: Switching between models you already know When to use ArchitectGBT: Figuring out which model to use in the first place
Think of it as research before execution. ArchitectGBT helps you make the decision, Cursor Auto helps you execute it.
Features
| Feature | Free | Pro | |---------|------|-----| | Browse 50+ AI Models | Unlimited | Unlimited | | AI Recommendations | 3/day | Unlimited | | Code Templates | No | Unlimited | | Cost Calculator | No | Yes | | API Keys | No | Up to 5 | | Priority Support | No | Yes |
AI Model Recommendations
Get intelligent, personalized suggestions for your specific use case:
- Smart analysis with detailed reasoning, pros/cons, and alternatives
- Budget optimization (low/medium/high/unlimited budgets)
- Priority matching (optimize for cost, speed, quality, or balanced)
- Real cost estimates based on realistic token usage
- Alternative suggestions if your top pick doesn't fit
Model Database (Always Free)
- 50+ AI models from OpenAI, Anthropic, Google, Meta, Mistral
- Real-time pricing per 1M tokens (input + output)
- Detailed specs (context windows, speed rankings, capabilities)
- Provider filtering to compare similar models
- Always updated with the latest model releases
Code Templates (Pro Only)
Production-ready integration code to ship faster:
- Copy-paste ready for Anthropic, OpenAI, Google Gemini
- TypeScript & Python support
- Complete examples with installation, env setup, and error handling
- Best practices including streaming, retries, and rate limiting
- Saves hours of reading API docs
View all features on ArchitectGBT →
Installation
Choose your editor below. No API key needed to start — upgrade to Pro later for unlimited access.
Cursor IDE
Cursor has native MCP support built-in.
Option 1: Free (No API Key)
- Create
.cursor/mcp.jsonin your project root:
{
"mcpServers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"]
}
}
}- Restart Cursor IDE
- Test it — Look for the connection icon in the bottom right, then ask:
- "Show me all available AI models"
- "Recommend an AI model for my chatbot project"
You're done. You get 3 free recommendations/day.
Option 2: Pro (Unlimited)
Upgrade to Pro → then:
- Get your API key from ArchitectGBT Dashboard
- Edit
.cursor/mcp.json:
{
"mcpServers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"],
"env": {
"ARCHITECTGBT_API_KEY": "agbt_your_key_here"
}
}
}
}- Replace
agbt_your_key_herewith your actual API key - Restart Cursor IDE
Unlimited recommendations + code templates unlocked.
Claude Desktop
Option 1: Free (No API Key)
Open config file:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
- Windows:
Add this configuration:
{
"mcpServers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"]
}
}
}- Restart Claude Desktop
- Test it — Ask Claude:
- "List all AI models from Anthropic"
- "Recommend a model for document analysis"
You're set. You get 3 free recommendations/day.
Option 2: Pro (Unlimited)
Upgrade to Pro → then:
- Get your API key from ArchitectGBT Dashboard
- Edit the config file (same locations as above):
{
"mcpServers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"],
"env": {
"ARCHITECTGBT_API_KEY": "agbt_your_key_here"
}
}
}
}- Replace
agbt_your_key_herewith your actual API key - Restart Claude Desktop
Unlimited access activated.
Other Editors
VS Code (with Continue extension)
- Install Continue extension
- Add to settings:
{
"mcp.servers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"]
}
}
}For Pro: Add "env": { "ARCHITECTGBT_API_KEY": "agbt_your_key" } after upgrading.
Zed Editor
Add to ~/.config/zed/settings.json:
{
"context_servers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"]
}
}
}For Pro: Add "env": { "ARCHITECTGBT_API_KEY": "agbt_your_key" } after upgrading.
Other MCP-Compatible Editors
Check ArchitectGBT Documentation for setup instructions.
Usage Guide
ArchitectGBT provides 3 MCP tools that work directly in your IDE:
1. list_models — Browse Models (Always Free)
Shows all 50+ AI models with pricing, specs, and capabilities.
Rate limit: Unlimited (no restrictions)
Example prompts:
"Show me all AI models"
"List models from Anthropic with pricing"
"What OpenAI models are available?"
"Compare Claude models"Response includes:
- Model name and provider
- Pricing per 1M tokens (input/output)
- Context window size
- Speed ranking
- Key capabilities
Tip: Start here to explore options before asking for a recommendation.
2. get_ai_recommendation — AI-Powered Suggestions
Analyzes your project and recommends the best model with detailed reasoning.
Rate limit:
- Free: 3 per day (resets at midnight UTC)
- Pro: Unlimited
Example prompts:
"Recommend an AI model for a customer support chatbot handling 10k requests/day"
"What's the best model for analyzing legal documents with 50k tokens?"
"I need a fast, cheap model for sentiment analysis — recommend one"
"Find me a model with low latency and medium budget for a translation app"Response includes:
- Top recommended model with match score
- Detailed reasoning (why it's perfect for your use case)
- Pros and cons
- Estimated daily cost
- Alternative suggestions
- Upgrade prompt if you hit the free limit
Important: This is the only rate-limited tool. Browse models (list_models) unlimited.
Hit your limit?
- Wait for daily reset (midnight UTC)
- Upgrade to Pro → for unlimited access
3. get_code_template — Production Code (Pro Only)
Returns production-ready integration code for specific AI models.
Rate limit:
- Free: Not available
- Pro: Unlimited
Example prompts:
"Give me TypeScript code to integrate Claude"
"Show me how to use GPT-4 in Python"
"Get me a Gemini integration template"
"I need example code for calling Anthropic's API with streaming"Response includes:
- Installation commands
- Environment variable setup
- Complete working code
- Usage examples with error handling
Not a Pro user? Free users see an upgrade prompt with pricing.
Get Pro access → to unlock code templates.
Prompting Tips
| Do | Don't |
|-------|----------|
| Be specific about your use case | Ask generic "which is best?" questions |
| Mention budget, volume, latency needs | Only ask for model names without context |
| Start with list_models to explore | Waste recommendations on exploratory questions |
| Use get_ai_recommendation for decisions | Ask for code without mentioning language |
Pro Tip: You can say "use the list_models tool" or "call get_ai_recommendation" to explicitly trigger a specific tool.
Pricing & Limits
| Tier | Recommendations | Code Templates | Cost Calculator | API Keys | Price | |------|-----------------|----------------|-----------------|----------|-------| | Free | 3/day (MCP) + 10/month (web) | No | No | No | $0 | | Pro | Unlimited | Unlimited | Yes | Up to 5 | $15/mo |
Free Tier
Perfect for trying ArchitectGBT:
- No signup required — works instantly
- 3 AI recommendations/day via MCP (IP-based rate limit)
- Unlimited model browsing (
list_modelstool) - 10 recommendations/month on the web app
- Full model database access on architectgbt.com
Limitations:
- No code templates
- No cost calculator
- No API key access
- No priority support
Best for: Exploring options, comparing models, trying ArchitectGBT for personal projects.
Pro Tier ($15/month)
For developers shipping to production:
- Unlimited AI recommendations (MCP + web app)
- Production-ready code templates (TypeScript + Python)
- Cost calculator with realistic usage estimates
- API key access (create up to 5 keys)
- Priority email support
- Advanced filters and comparison tools
- Early access to new features
Best for: Professional developers, startups, teams building AI products.
Enterprise (Custom Pricing)
Need more? We offer:
- Team collaboration (shared recommendations, templates)
- SSO and advanced security
- Priority support with SLA
- Onboarding and training
Rate Limit Details
| Endpoint | Free | Pro |
|----------|------|-----|
| list_models | Unlimited | Unlimited |
| get_ai_recommendation (MCP) | 3/day | Unlimited |
| Web app recommendations | 10/month | Unlimited |
| get_code_template | No | Unlimited |
Rate limit resets: Midnight UTC daily
Hit your limit?
- Wait for reset (resets every 24 hours)
- Use the web app (10 free/month)
- Upgrade to Pro for unlimited access
Available Tools
This MCP server exposes 3 tools that your IDE's AI assistant can use:
1. list_models
Description: Browse all available AI models with pricing and specifications.
Parameters:
provider(optional): Filter by provider —"OpenAI"|"Anthropic"|"Google"|"Meta"|"Mistral"limit(optional): Max models to return (default: 50)
Returns:
- List of models with name, provider, pricing (input/output per 1M tokens), context window, speed ranking
Rate limit: Unlimited (always free)
2. get_ai_recommendation
Description: Get AI-powered model recommendations based on your project requirements.
Parameters:
projectDescription(required): Your use case (e.g., "customer support chatbot")budget(optional):"low"|"medium"|"high"|"unlimited"(default: medium)priority(optional):"cost"|"speed"|"quality"|"balanced"(default: balanced)
Returns:
- Top recommended model with match score
- Detailed reasoning and analysis
- Pros and cons
- Estimated costs
- Alternative suggestions
Rate limit:
- Free: 3 per day (resets midnight UTC)
- Pro: Unlimited
3. get_code_template
Description: Get production-ready integration code for specific AI models.
Parameters:
model(required): Model name (e.g.,"Claude","GPT-4","Gemini")language(optional):"typescript"|"python"(default: typescript)
Returns:
- Installation commands
- Environment variable setup
- Complete working code
- Usage examples with error handling
Rate limit:
- Free: Not available (shows upgrade prompt)
- Pro: Unlimited
Troubleshooting
"Daily Limit Reached"
Problem: You've used your 3 free recommendations for today.
Solutions:
- Wait for reset — Limits reset at midnight UTC (check countdown in response)
- Use the web app — Get 10 more recommendations/month at architectgbt.com
- Upgrade to Pro — Get unlimited access for $15/month → Upgrade here
"API Key Invalid"
Problem: Your API key isn't being accepted.
Solutions:
- Check format — Keys start with
agbt_(32 characters total) - Verify Pro status — API keys only work for Pro subscribers → Check subscription
- Regenerate key — Create a new key at Settings
- Check config — Ensure key is in
envsection of your MCP config - Restart IDE — Changes require full restart
Example config:
{
"mcpServers": {
"architectgbt": {
"command": "npx",
"args": ["-y", "architectgbt-mcp@latest"],
"env": {
"ARCHITECTGBT_API_KEY": "agbt_abc123xyz..."
}
}
}
}MCP Server Not Loading
Problem: Tools aren't showing up in your IDE.
Solutions:
- Check Node.js — Requires Node.js >= 18.0.0 (
node --version) - Verify config — Check JSON syntax in your MCP config file
- Restart IDE — Full restart required after config changes
- Test manually — Run
npx -y architectgbt-mcp@latestin terminal - Check network — Verify you can access https://architectgbt.com/api/models
For Cursor users: Look for the connection icon in the bottom right corner
Models Not Showing
Problem: list_models returns empty or errors.
Solutions:
- Check internet — Test connection to https://architectgbt.com
- Verify API — Visit https://architectgbt.com/api/models in browser
- Update package — Run with
@latesttag:npx -y architectgbt-mcp@latest - Clear cache — Delete
node_modules/.cacheand retry
Free Tier Not Working
Problem: Getting errors even without API key.
Solutions:
- Remove API key — Free tier doesn't need
envsection at all - Use basic config:
{ "mcpServers": { "architectgbt": { "command": "npx", "args": ["-y", "architectgbt-mcp@latest"] } } } - Check IP limit — Each IP gets 3/day (shared across users on same network)
- Create free account — Get 10 more/month on architectgbt.com
Still Having Issues?
- Email support: [email protected]
- Documentation: architectgbt.com/docs/mcp-integration
- Report bugs: GitHub Issues
Environment Variables
| Variable | Required | Description |
|----------|----------|-------------|
| ARCHITECTGBT_API_KEY | No (Pro only) | Your API key for unlimited access (starts with agbt_) |
| ARCHITECTGBT_API_URL | No | Custom API endpoint (default: https://architectgbt.com) |
Example:
export ARCHITECTGBT_API_KEY="agbt_your_key_here"Links & Support
ArchitectGBT Platform
- Homepage: architectgbt.com — Explore 50+ models, compare pricing, get recommendations
- Browse Models: architectgbt.com/models — Full database with filters and search
- Pricing: architectgbt.com/pricing — Upgrade to Pro for unlimited access
- Get API Key: architectgbt.com/dashboard/settings — Manage keys and subscription
- Documentation: architectgbt.com/docs/mcp-integration — Full MCP setup guide
- Blog: architectgbt.com/blog — Model comparisons, benchmarks, tutorials
NPM Package
- Package: npmjs.com/package/architectgbt-mcp
- Version: 0.4.0
- License: MIT
Support Channels
- Email: [email protected]
- Bug Reports: GitHub Issues
- Feature Requests: GitHub Discussions
- Twitter/X: @architectgbt — Updates and announcements
Why Use ArchitectGBT?
Save hours of research — We track 50+ models from 5 providers so you don't have to.
Make better decisions — AI-powered recommendations based on your actual requirements, not marketing hype.
Ship faster — Production-ready code templates eliminate boilerplate and integration headaches.
Optimize costs — Realistic cost estimates prevent budget surprises in production.
Stay updated — We monitor pricing changes, new releases, and deprecations across all providers.
Contributing
This MCP server is part of the ArchitectGBT platform. We welcome contributions.
Development Setup:
git clone https://github.com/3rdbrain/architectgbt-mcp.git
cd architectgbt-mcp
npm install
npm run build
npm run devBefore submitting PRs:
- Test with both Cursor and Claude Desktop
- Update documentation if adding features
- Follow existing code style
License
MIT © ArchitectGBT
Acknowledgments
Built with:
- Model Context Protocol (MCP) by Anthropic
- Cursor IDE and Claude Desktop
- TypeScript, Node.js, and the amazing open-source community
Built for developers who ship fast
