An agentic platform for earth intelligence: Earth Virtual Expert’s community-driven Model Context Protocol registry

  • * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

  • Date
    1 December 2026
    Timeframe
    14:00 - 15:30 CEST, Geneva
    Duration
    1h 30 minutes
    • Days
      Hours
      Min
      Sec

    Answering an Earth Observation (EO) question often means jumping between disaster alerts, satellite catalogues, scientific papers and processing services, each with its own portal and format. Artificial Intelligence (AI) agents can now take over that orchestration, but only if the tools behind them are open, trustworthy and built for EO.

    This AI for Good workshop introduces ESA Φ-lab’s EVE (Earth Virtual Expert), a fully open-source agentic platform for Earth Observation developed by Pi School. EVE combines EVE-Instruct, an open-weight EO-adapted model built together with Mistral AI, retrieval over peer-reviewed literature with citations, and a community-curated registry of EO tools exposed through the Model Context Protocol (MCP), an open standard that lets AI agents call external services.

    Following a disaster-response scenario, participants will use a public API key and a ready-made notebook to query EVE, call tools like GDACS disaster alerts, or EFFIS wildfire data, and see how the agent chains them into a traceable, cited answer. We will then plug the same tools into a LangGraph agent and show how anyone can contribute tools and skills to the registry.

    No AI engineering background is needed: the session is built for EO practitioners who want to put agents to work.

    Workshop objectives

    By the end, participants will be able to:

    • query EVE through its API and Model Context Protocol (MCP) endpoint
    • call EO tools (GDACS, EFFIS) and read the agent’s trace
    • plug EVE tools into their own LangGraph agent
    • and contribute new tools and skills to the open registry.

    Are you sure you want to remove this speaker?