Course

Building GeoAI Agents for Fun & Profit

Shoaib Burq

Geospatial Software Engineer

Learn to build AI Agents for location-based intelligence

As we build ever more sophisticated agentic applications, where does location data sit? New protocols, standards, and UX patterns have emerged — many built for text-first apps. This course tackles what changes when maps, routing APIs, and geocoding join the agent stack: MCP tool servers, AG-UI chat UIs, multi-agent A2A, and context engineering so heavy GeoJSON never bloats the model. Hands-on labs use OpenRouteService, Pydantic AI, CopilotKit, and MapLibre — in Colab or locally.

What you’ll learn

Ship map-and-chat GeoAI apps: MCP geospatial tools, Pydantic AI agents, and AG-UI UIs, with context engineering for GeoJSON.

  • Implement FastMCP tools over OpenRouteService routing with typed parameters and small JSON results the model can read.

  • Run a detached SSE MCP server on :8004 and connect hosts (Colab, Cursor, pydantic-ai) via MCPToolset.

  • Smoke-test tools in a notebook with await before wiring an agent, same logic as the stdio server.

  • Attach MCP toolsets to a Pydantic AI agent so the model plans calls and reads compact tool results.

  • Expose the agent to the browser over AG-UI and stream replies through a CopilotKit chat UI.

  • Swap LLM providers via configuration without rebuilding the agent stack.

  • Use AG-UI client tools so the browser draws routes while the model only sees summaries.

  • Apply context engineering: thin MCP tool outputs to the LLM, full geometry via a separate channel.

  • Deliver distance, duration, and a visible map—not raw GeoJSON in the conversation.

  • Resolve landmarks and addresses to coordinates through MCP geocoding before calling routing tools.

  • Support natural-language prompts like “drive from the station to the square” end to end.

  • Handle ambiguous locations by confirming with the user instead of guessing.

  • Use an orchestrator agent over AG-UI that delegates routing and weather to A2A specialist agents.

  • Keep typed summaries in the orchestrator’s context while geometry stays in specialist-side caches.

  • Combine trip routing with destination context—travel time plus conditions at arrival.

  • Start with in-process tools to learn the agent loop; adopt MCP when multiple apps need the same catalog.

  • Use A2A when a whole domain—credentials, retries, policy—should sit behind another agent.

  • Leave with a clear map of which protocol handles tools, user chat, and agent-to-agent work.

Learn directly from Shoaib

Shoaib Burq

Shoaib Burq

Contact

Shoaib Burq · CTO, Decision Labs · Geobase · agentic geospatial AI instructor

Geoscience Australia
European Space Agency (ESA)
Geoscience Australia
Hugging Face
Morgan Stanley
See all products from Shoaib Burq

Who this course is for

  • Developers building location-aware apps who want MCP tools, AG-UI chat, and maps—without dumping raw geometry into the model.

  • GIS and geospatial analysts who know maps and APIs and want a practical path to conversational assistants users can trust.

  • Tech leads evaluating MCP, AG-UI, and A2A for geo workflows—when to share tools, split specialists, and keep models focused.

What's included

Lifetime access

Go back to course content and recordings whenever you need to.

Community of peers

Stay accountable and share insights with like-minded professionals.

Certificate of completion

Share your new skills with your employer or on LinkedIn.

Maven Guarantee

Your purchase is backed by the Maven Guarantee.

Course syllabus

8 live sessions • 19 lessons • 4 projects

Week 1

Jul 13—Jul 19

    Before you start: Set up your geo agent stack

    • Jul

      14

      Setup, troubleshooting and Q&A

      Tue 7/145:00 PM—6:00 PM (UTC)
    5 more items • Free preview

    Week 1 — MCP: tool calling

    • Jul

      18

      Live 1: MCP tools + Exercise 1

      Sat 7/183:00 PM—5:30 PM (UTC)
    4 more items

    Week 1 — AG-UI: chat and map

    • Jul

      18

      Live 1b: AG-UI + routing map (Exercises 2 and 3)

      Sat 7/186:30 PM—9:00 PM (UTC)
    3 more items

Week 2

Jul 20—Jul 26

    Week 2 — MCP: multi-tool servers

    • Jul

      22

      Office hour: Week 1 labs + Project 1 (Americas)

      Wed 7/225:00 PM—6:00 PM (UTC)

    • Jul

      22

      Optional: Office hour: Week 1 (EU / Africa)

      Wed 7/222:00 PM—3:00 PM (UTC)
      Optional

    • Jul

      25

      Live 2: Multi-tool MCP + AG-UI (Exercise 5)

      Sat 7/253:00 PM—5:30 PM (UTC)
    4 more items

    Week 2 — A2A: orchestration

    • Jul

      25

      Live 2b: A2A orchestration + capstone kickoff (Exercise 6)

      Sat 7/256:30 PM—9:00 PM (UTC)
    4 more items

Frequently asked questions

Maven for Teams

Reimbursement

Get your company to pay

Everything L&D needs: email template, receipts, and certificate of completion.

Get reimbursed

Private cohort

Run a cohort for your org

A dedicated cohort with a custom schedule and curriculum, tailored to your team.

Book a private cohort

Get course updates