A minimal example using Agent Development Kit showcasing a simple weather and time agent that can call multiple tools.
If you have not already done so, follow the instructions in the Install Guide to create the development environment and install NeMo Agent Toolkit.
From the root directory of the NAT library, run the following command:
uv pip install -e examples/frameworks/adk_demoFor this example, an OpenAI API key is required. You can set it as follows:
export OPENAI_API_KEY="<your_openai_key>"
# Optional (defaults to https://api.openai.com/v1 if unset)
export OPENAI_API_BASE="<your_openai_base_url>"Google ADK support within NeMo Agent toolkit currently only supports OpenAI and Azure OpenAI models for tool calling.
Run the workflow with the NAT CLI
nat run --config_file examples/frameworks/adk_demo/configs/config.yml --input "What is the weather and time in New York today?"<snipped for brevity>
Configuration Summary:
--------------------
Workflow Type: adk
Number of Functions: 2
Number of Function Groups: 0
Number of LLMs: 1
Number of Embedders: 0
Number of Memory: 0
Number of Object Stores: 0
Number of Retrievers: 0
Number of TTC Strategies: 0
Number of Authentication Providers: 0
<snipped for brevity>
--------------------------------------------------
Workflow Result:
['Here’s the latest for New York:\n- Weather: Sunny, around 25°C (77°F)\n- Time: 2025-09-25 12:27:26 EDT (UTC-4)']
--------------------------------------------------