Freeform Tool Calling with GPT-5 in Azure AI Foundry
April from Microsoft Developer demonstrates how to use freeform tool calling with GPT-5 in Azure AI Foundry, showing a seamless workflow from natural language to code, data artifacts, and visualization.
Freeform Tool Calling with GPT-5 in Azure AI Foundry
In this episode, April from Microsoft Developer explores the latest freeform tool calling feature in GPT-5, available through Azure AI Foundry. This new approach enables the model to send raw text payloads—including Python scripts, SQL queries, and configuration files—directly to external tools, bypassing the need for structured JSON wrappers and rigid schemas.
Key Highlights
Send Raw Text Payloads:
- GPT-5 can now generate and transmit Python, SQL, and config files as plain text to external tools.
- Eliminates integration overhead caused by requiring structured JSON schemas, making AI-powered automation more flexible.
Workflow Demonstrated:
- April shows how GPT-5 generates SQL code to compute results from a data source.
- The output is routed to Python scripts for further formatting and visualization.
- The workflow returns tangible, reproducible artifacts such as CSV files and charts, all initiated via natural language prompts.
Hands-On Demos:
- Step-by-step code review, execution, and output inspection.
- Examples provided for both SQL computations and Python-powered visualizations.
Getting Started:
- Try freeform tool calling yourself at ai.azure.com
- Access source code and further resources:
Video Chapters
- 0:00 - 1:47: Welcome and introduction
- 1:48 - 7:48: Code review
- 7:49 - 12:45: Run the script & review output
- 12:46 - 13:14: Summary and more
Why This Matters
Freeform tool calling in Azure AI Foundry and GPT-5 streamlines integration for developers and technical users, offering a more flexible, natural language-driven interface for orchestrating complex workflows—code generation, data processing, and artifact creation—on Microsoft’s AI cloud platform.