Skip to main content
The fastest way to learn GroundForge is to complete one small weather example and then follow the same request in the platform.

Before you begin

You need:
  • access to platform.groundforge.ai;
  • an Organization and Workspace;
  • the gforge CLI for command-line steps;
  • Python 3.10 or newer;
  • the groundforge-langchain wheel supplied with your GroundForge release for the Agent example.

Path A: Weather MCP

modelcontextprotocol/quickstart-resources

Official Python MCP server with get_alerts(state) and get_forecast(latitude, longitude).
Follow Weather MCP Example to discover the tools, create Gateway rules, test an allowed call, and review the resulting records.

Path B: LangChain weather Agent

groundforgeai/demo

GroundForge external-Agent examples using automatic Channel input handling and explicit adapters.
Choose one: Then attach a Channel, send a weather question, apply a Gateway rule, and open the Trace.

What to check in GroundForge

After the first request, confirm:
  1. the Agent or MCP provider is visible in the selected Workspace;
  2. Discovery lists the expected weather capability;
  3. the Gateway shows the action you selected;
  4. an Operator Task appears when the action is Review;
  5. the Trace shows the Channel input, decision, capability call, and response.
The next concept page explains how those resources relate: Resource Model.
Last modified on July 27, 2026