GeoScout AI

• 

  chat interface

• 

  index calculation

🎯 The Problem

• $8 trillion in critical minerals needed by 2050
• US imports 80% of rare earths from China - national security risk
• Billions of satellite images archived but underutilized
• Traditional mineral exploration: slow, expensive, risky

💡 Our Solution: GeoScout AI AI-powered mineral exploration platform that unlocks dormant Earth observation data using:

• Real satellite imagery (Sentinel-2, 10m resolution)
• MCP architecture (Model Context Protocol from Anthropic)
• Claude AI reasoning for geological interpretation
• LangGraph agents for intelligent analysis

🏗️ Architecture Satellite Data (Sentinel-2) ↓ MCP Server (exposes data as queryable tools) ↓ LangGraph Agent (orchestrates analysis) ↓ Claude AI (geological reasoning) ↓ Streamlit UI (interactive exploration)

🛰️ Live Demo Sites

• Mountain Pass, CA - Rare Earth Elements (carbonatite, 15% of global supply)
• Bingham Canyon, UT - Copper-Molybdenum (largest porphyry deposit)
• Bokan Mountain, AK - REE + Uranium (peralkaline granite)

All with 2024 imagery, 0% cloud cover

🔬 What It Does Spectral Mineral Indices:

• Iron Oxide Index (Red/Blue ratio) → Oxidized minerals, gossans
• Clay Alteration Index (SWIR bands) → Hydrothermal alteration zones
• Ferrous Index → Iron-bearing minerals
• NDVI → Vegetation masking

AI Capabilities:

• Natural language queries about mineral potential
• Automated spectral analysis
• Exploration target recommendations
• Multi-site comparative analysis

🚀 Why It Wins Technical Innovation:

✅ First geospatial application of MCP protocol ✅ Real satellite data, not mock/API wrapper ✅ Production-quality code (2,197 lines, fully functional) ✅ Secure data serving (MCP tool-level authentication)

Strategic Impact:

✅ Unlocks billions in archived satellite data ✅ Addresses critical mineral supply chain security ✅ Scalable to global coverage (any satellite sensor) ✅ Reduces exploration costs and timelines

Built in 3 hours. Cost: $2.28

📊 Key Metrics

• 3 sites analyzed with real 2024 Sentinel-2 data
• 6 spectral bands per scene (Blue, Green, Red, NIR, SWIR1, SWIR2)
• 30m spatial resolution (10m for visible/NIR)
• 4 specialized tools in MCP server
• Claude Sonnet 4 for geological reasoning

🎓 The Vision

• Today: 3 sites, Sentinel-2 data
• Tomorrow: Global coverage, multi-sensor fusion
• Future: Real-time mineral discovery at planetary scale
• Transform Earth observation from passive archive to active intelligence.

Built with Claude Code + Anthropic Claude AI Team: ClaAko (Claude + Ako) + Claude Code GitHub: https://github.com/ako1983/geoscout

Built With

• claude
• langchain
• langgrapgh
• mcp
• python