Skip to main content

groundforgeai/demo

The simple_agent_service.py example creates a LangChain Agent with a safe demo tool: get_weather(city).

1. Create the external Agent in GroundForge

In platform.groundforge.ai:
  1. select the intended Organization and Workspace;
  2. create an External Agent;
  3. create a Service Account for the Agent process;
  4. issue an Agent-targeted credential;
  5. keep the stable external Agent ID, Organization ID, and Workspace ID.
Screenshot placeholder
Add /images/examples/external-agent-create.png showing the External Agent form and runtime identity.

2. Clone and install

Use the SDK wheel supplied with your GroundForge release.

3. Set the connection values

The demo does not require custom caller metadata. GroundForge sends the normalized Channel input, and automatic mode converts it into the Agent’s message input.

4. Run the Agent

The example uses:
Keep the process running. In GroundForge, confirm that the Agent is connected and Discovery lists get_weather(city).
Screenshot placeholder
Add /images/examples/langchain-agent-connected.png showing connection state and get_weather in Discovery.

5. Attach a Channel

Choose Twilio WhatsApp, Mailgun, Slack, or a JSON webhook. Attach the Channel to the Agent and publish the Agent before configuring the provider callback.

6. Send a request

GroundForge sends the normalized request to the running Agent. The Agent selects get_weather(city="Ottawa"), the Gateway checks the rule, and the final response returns through the Channel.

7. Govern and trace the tool

Set get_weather to Observe, Review, or Block and send the same question again. Open the Trace to compare the paths. Continue with LangChain Agent with Adapters, Govern the Weather Capability, and Request Tracing.
Last modified on July 27, 2026