llm-radar-mcp 0.1.0

MCP server providing real-time AI model intelligence - pricing, capabilities, and recommendations

LLM Radar

Real-time AI Model Intelligence via MCP

Skip the search. Your AI already has current model info.

Live Dashboard · Model Reference · MCP Setup · Contributing

---

What is LLM Radar?

LLM Radar is an MCP server that gives your AI assistant current information about AI models from OpenAI, Anthropic, and Google.

The problem: AI assistants have training cutoffs. Ask about models and you get outdated recommendations, deprecated APIs, or hallucinated pricing.

The solution: Connect LLM Radar and your AI already knows what's available today:

• Fetching fresh data from provider APIs daily
• Enriching it with Claude for better descriptions
• Exposing it via MCP for any compatible client

---

MCP Server Setup

Option 1: Remote Server (Recommended)

Connect directly to the hosted MCP server - no installation needed:

Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "llm-radar": {
      "url": "https://ajents.company/llm-radar/mcp"
    }
  }
}

Option 2: Local via npx/pip

# Install
pip install llm-radar-mcp

# Or run directly
pip install llm-radar-mcp && llm-radar-mcp

Claude Desktop config (local stdio):

{
  "mcpServers": {
    "llm-radar": {
      "command": "llm-radar-mcp"
    }
  }
}

Option 3: Docker

docker run -p 8000:8000 ghcr.io/ajentsor/llm-radar:latest

Then connect to http://localhost:8000/sse

---

Available MCP Tools

Once connected, you can use these tools:

Tool	Description
query_models	Search/filter models by provider, type, or modality support
compare_models	Side-by-side comparison of specific models
get_model	Get detailed info about a specific model by API ID
list_model_ids	List all available model IDs for a provider

Example Queries

"What models support vision input?"
→ Uses query_models with input_modality="image"

"Compare GPT-4o, Claude Sonnet, and Gemini 2.5 Pro"
→ Uses compare_models with those model IDs

"List all OpenAI model IDs"
→ Uses list_model_ids with provider="openai"

---

Available Resources

The MCP server also exposes resources you can read directly:

Resource URI	Description
llm-radar://models/all	Complete JSON data
llm-radar://models/openai	OpenAI models only
llm-radar://models/anthropic	Anthropic models only
llm-radar://models/google	Google models only
llm-radar://highlights	Curated recommendations

---

How It Works

┌─────────────────────────────────────────────────────────────────┐
│                    Daily GitHub Action (8am UTC)                │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  ┌──────────┐   ┌──────────┐   ┌──────────┐                    │
│  │  OpenAI  │   │Anthropic │   │  Google  │   ← Fetch APIs     │
│  │   API    │   │   API    │   │   API    │                    │
│  └────┬─────┘   └────┬─────┘   └────┬─────┘                    │
│       │              │              │                           │
│       └──────────────┼──────────────┘                           │
│                      ▼                                          │
│              ┌──────────────┐                                   │
│              │    Claude    │   ← Enrich & Format               │
│              │   (Sonnet)   │                                   │
│              └──────┬───────┘                                   │
│                     │                                           │
│       ┌─────────────┼─────────────┐                            │
│       ▼             ▼             ▼                            │
│  ┌─────────┐  ┌──────────┐  ┌───────────┐                      │
│  │models.  │  │ MCP      │  │  GitHub   │   ← Deploy           │
│  │  json   │  │ Server   │  │  Pages    │                      │
│  └─────────┘  └──────────┘  └───────────┘                      │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

---

Data Format

Each model includes:

Field	Description
id	API model identifier
name	Human-friendly name
provider	openai, anthropic, or google
description	What the model is best for
context_window	Max input tokens
pricing	Input/output cost per 1M tokens
capabilities	vision, function_calling, reasoning, etc.
status	active, preview, or deprecated
released	Release date
recommended_for	Use case suggestions

---

Local Development

# Clone
git clone https://github.com/ajentsor/llm-radar.git
cd llm-radar

# Install
python3 -m venv venv
source venv/bin/activate
pip install -e ".[dev]"

# Run MCP server (stdio mode)
llm-radar-mcp

# Run MCP server (HTTP mode for testing)
llm-radar-mcp --http --port 8000

# Fetch fresh data (requires API keys)
cp .env.example .env
# Edit .env with your API keys
python3 -m llm_radar.fetch_models
python3 -m llm_radar.aggregate_with_claude

---

Project Structure

llm-radar/
├── src/llm_radar/              # Main package
│   ├── __init__.py
│   ├── mcp_server.py           # MCP server implementation
│   ├── fetch_models.py         # API fetchers
│   └── aggregate_with_claude.py # Claude enrichment
├── data/
│   ├── models.json             # Structured model data
│   ├── MODELS.md               # Human-readable reference
│   └── raw/                    # Raw API responses
├── docs/                       # GitHub Pages site
├── Dockerfile                  # Container build
├── docker-compose.yml          # Local container setup
├── pyproject.toml              # Python package config
└── .github/workflows/
    └── update-models.yml       # Daily cron job

---

Configuration

To run the data fetcher yourself:

# .env file
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AI...

For GitHub Actions, add these as repository secrets.

---

Self-Hosting

Docker Compose

version: '3.8'
services:
  llm-radar:
    image: ghcr.io/ajentsor/llm-radar:latest
    ports:
      - "8000:8000"
    restart: unless-stopped

Cloudflare Workers / Fly.io / Railway

The MCP server supports HTTP/SSE transport, making it deployable to any platform that supports long-running HTTP connections.

---

Contributing

See CONTRIBUTING.md for guidelines.

Key areas for contribution:

• Additional providers (Cohere, Mistral, etc.)
• More MCP tools
• Better data enrichment prompts
• Documentation improvements

---

License

MIT License - see LICENSE

---

Built for developers who want accurate AI model info

Star this repo · Report Issue · View Dashboard