microsoft/Planetary-Explorer

microsoft/Planetary-Explorer

Description: An AI powered geospatial application that allows you to explore and visualize Earth science data using natural language.

Language: Python

License: MIT

Stars: 171

Forks: 46

Open issues: 1

Created: 2025-09-15T16:46:22Z

Pushed: 2026-06-19T03:09:27Z

Default branch: main

Fork: no

Archived: no

README:

🌍 Welcome to Planetary Explorer!

Planetary Explorer, built on AI Foundry, demonstrates how organizations can use Microsoft Planetary Computer Pro to combine geospatial data with generative AI experiences. By enabling users to explore Earth science data through natural language, it makes complex geospatial workflows more accessible to analysts, operators, and decision makers—not just GIS specialists. This helps teams accelerate insight generation and support scenarios ranging from operational monitoring to risk management.

📋 Overview

Planetary Explorer turns natural-language questions into grounded geospatial answers. Its multi-agent system picks the right data, renders it on the map, and reasons over the result.

It fuses mutliple surfaces behind one chat:

• Microsoft Planetary Computer — 130+ public STAC collections & MPC Pro / GeoCatalog in your tenant for private collections
• Microsoft Fabric Lakehouse — delta tables and compute feed workflows
• Azure AI Search — documentation for grounding responses
• Foundry LLMs, weather + geospatial models — GPT, Aurora, NVIDIA Earth-2 FCN, and MAI Weather

Meet users where they already work:

• React web app — purpose-built map + chat experience
• Microsoft Teams — chat with Planetary Explorer agents in any channel
• M365 Copilot — declarative agent surfaces the same answers inside Word, Outlook, and Copilot Chat
• VS Code / Claude Desktop — every agent exposed as MCP tools for developers

Built on Microsoft Agent Framework, Azure AI Agent Service, Semantic Kernel and Model Context Protocol so analysts, operators, and decision-makers spend less time wrangling data and more time acting on insight.

Watch Satya Nadella introduce NASA Earth Copilot, the inspiration behind Planetary Explorer, at Microsoft Ignite 2024: View Here

Auto-Deploy Ready: This repository includes fully automated deployment via Bicep and GitHub Actions. Follow the [Quick Start Guide](QUICK_DEPLOY.md) to deploy the complete architecture: infrastructure, backend, and frontend within one hour. Its modular architecture is designed for extensibility.

> Planetary Explorer is a reusable geospatial AI pattern that can be adapted across different use cases. It is not a supported Microsoft product.


✨ Features

• Multi-Agent Architecture — Microsoft Agent Framework (MAF) workflows + Semantic Kernel + Azure AI Agent Service.
• Dual MPC Surface — Chat over MPC Public *or* MPC Pro / GeoCatalog in your own tenant
• Pluggable Connection Surfaces — Bring your own Microsoft Fabric Lakehouse, Azure AI Search indexes, and Foundry geospatial + weather models.
• MCP Server — Expose every agent as Model Context Protocol tools for VS Code GitHub Copilot, Claude Desktop, and other MCP clients.
• Multiple Client Surfaces — One backend, your choice of UI: a purpose-built React web app, a Microsoft Teams bot, or an M365 Copilot declarative agent.
• Copilot Studio & ArcGIS — Custom connectors for Copilot Studio, plus optional Esri ArcGIS integration for enterprise GIS workflows.
• Fully Private Deployment — Optional VNet integration with private endpoints, private DNS zones, and Entra ID authentication for an enterprise-ready deployment out of the box.


🎯 Use Cases

| | | | | | |:---:|:---:|:---:|:---:|:---:| | Science & Environment | Agriculture & Natural Resources | Energy & Infrastructure | Public Safety & Emergency Management | Defense / National Security | | Accelerates climate, air quality, land-surface, extreme weather scenarios, and environmental research | Assess drought conditions, soil moisture, and water quality for agriculture planning | Monitor energy grids, transmission corridors, and dam infrastructure, supporting site selection and permitting | Supports response to wildfires, floods, hurricanes, and other natural disasters | Monitor geospatial intelligence and support situational awareness for national security operations |

🛰️ What Planetary Explorer Does


Query Examples

STAC Agent — chat-to-map (MPC Public + MPC Pro)

| Query | |-------| | Show coastal land cover changes in California | | Show me Sentinel-2 imagery over Los Angeles on May 20, 2026 | | Show me radar imagery of Houston Texas during Hurricane Harvey August 2017 |

Flip the MPC Pro toggle in the UI and every STAC query now runs against your tenant's collections.

Raster Sampling + Contextual Agent

| Action | Query | |--------|-------| | Pin drop → Chat | Sample the raster value at this location | | Chat | How do I interpret this collection? | | Chat | Explain what each class in this land-cover raster means |

GEOINT Modules — Vision, Terrain, Mobility, Comparison, Building Damage

| Module | Query | |--------|-------| | Vision | Analyze this satellite image — what land cover types are visible and what is the surface reflectance? | | Terrain | Is this location suitable for a construction permit? Analyze slope, flood risk, and flat areas. | | Terrain | Analyze terrain elevation, slope, and line-of-sight at 38.9N, 77.0W | | Comparison | Show wildfire activity in Southern California in January 2025 and analyze how it evolved over 48 hours | | Mobility | Classify terrain traversability at these coordinates across 5 elevation layers | | Building Damage | Assess building damage using before/after satellite imagery at these coordinates |

Extreme Weather Agent — NASA NEX-GDDP-CMIP6 (NetCDF + trend reasoning)

| Query | |-------| | What is the projected annual precipitation and peak daily rainfall for New Orleans? | | Compute the precipitation trend for New Orleans from 2020 to 2080 | | What are the projected temperature and precipitation trends for...