Tools are the foundation of agent AI capabilities. Just as we rely on specialized software like FEA solvers, search engines, and databases to interact with our work environments, AI agents need tools to extend their reach beyond into a border tooling environment. In this latest update, we’ve opened tool creation to users, enabling seamless integration with your existing knowledge bases and workflows.

What Are Toolboxes?

Toolboxes are collections of executable code that agents can run autonomously. Each toolbox contains a set of individual tools—think of them as the hammer and wrench in a toolbelt—which let agents access the world. A tool could be an API call to an external service or perhaps an interface with your local custom database. When an agent uses a tool, it receives structured information back that then informs its design or analysis process.

Our tooling system builds on the Model Context Protocol (MCP) framework, with adaptations optimized for Davinci’s Python browser environment. This means you can export your toolboxes to standard MCP-formatted Python for use in other environments, ensuring your investment in custom tooling remains portable.

How to Create and Use Toolboxes

Setting up a toolbox is straightforward:

1. Define the toolbox object in your model with a clear name and description
2. Add your functions (the individual tools) with proper documentation
3. Test the toolbox to verify it returns the expected data and behaves correctly
4. Enable it as active to make it available to all agents in your workspace

Once activated, any new agent will automatically have access to your toolbox.

Toolboxes vs. Standard Code Execution

While Davinci already supports Python script execution in the browser, toolboxes offer distinct advantages in that they are instantly accessible to agents without requiring user prompts or searches through your codebase. This creates a more direct path from agent reasoning to tool execution.

Expanding Davinci’s Capabilities

Custom toolboxes serve as the bridge between Davinci and your specialized work environment. For example, you can:

• Access external APIs like Digi-Key to access component databases or supplier information
• Connect to analysis servers for compute-intensive calculations
• Integrate proprietary databases that hold your organization’s specific data

This capability represents a significant leap forward in making Davinci truly adaptable to your unique engineering and design requirements.

The Future of Extensible AI

By opening up tool creation to users, we’re enabling a new level of customization that goes far beyond simple prompt engineering. Your agents can now interact with the same tools and data sources you use daily, creating truly integrated AI-assisted workflows. We’re excited to see how users will extend Davinci’s capabilities through toolboxes.