Languages: English (en)

Agent Tool Description Format (ATDF)

Welcome to the documentation for the Agent Tool Description Format (ATDF), an open protocol for describing tools functionally to enable AI agents to select and use them based on purpose, context, and operation, without relying on specific implementation details.

Current Version

Current version: 0.2.0 - See the changelog for details on the latest updates.

Introduction

ATDF is designed to solve the problem of tool integration for AI agents. Instead of requiring hard-coded tool names or complex technical APIs, ATDF provides a standardized way to describe tools based on:

What the tool does
When it should be used
How it is used

This functional approach allows AI agents to select tools based on the task at hand, rather than requiring specific knowledge about tool names or APIs.

Key Features

Core Features (v0.1.0)

Simple, Human-Readable Format: JSON/YAML structure that’s easy to understand.
Model Agnostic: Works with any AI agent model.
Tool Agnostic: Describes both physical tools (e.g., drills) and digital tools (e.g., APIs).
Prompt Agnostic: Tool selection based on function, not specific names.
Schema Validation: JSON Schema for validating tool descriptions.

Enhanced Features (v0.2.0)

Metadata Support: Organize tools with version, author, tags, and category information.
Rich Multilingual Support: Built-in localization for multiple languages.
Prerequisites and Dependencies: Specify required tools, conditions, and permissions.
Feedback Mechanisms: Progress indicators and completion signals.
Usage Examples: Real-world examples with inputs and expected outputs.
Complex Input Types: Support for nested objects and advanced schemas.

Quick Links

Specification: Detailed technical specification of the ATDF protocol.
Examples: Sample tool descriptions and how to create your own.
Multilingual Support: Information about using multiple languages.
Contributing: Guidelines for contributing to ATDF.
Changelog: History of ATDF versions and changes.

Getting Started

To start using ATDF, you can:

Explore Examples: Check out the example tool descriptions to understand the format.
Create Your Own: Follow the specification to create tool descriptions.
Validate Tools: Use the validator to ensure your tool descriptions are valid:
```
python tools/validator.py path/to/your/tool.json
```
Try the Demo: Run the demonstration agents to see ATDF in action:
```
python tools/demo/atdf_showcase.py
```

Use Cases

ATDF is designed for a wide range of applications, including:

AI Agents: Helping AI systems select and use tools appropriately.
Robotics: Describing physical tools for robotic systems.
API Integration: Standardizing descriptions of APIs and web services.
Multimodal Applications: Bridging different types of tools in a unified format.
Multilingual Systems: Supporting tool descriptions across different languages.

License

ATDF is licensed under the MIT License. See the LICENSE file for details.

Core Components

Specification: The core technical details of the format
Examples: Sample tool descriptions
Usage Examples: Practical examples of using ATDF
Multilingual Support: How localization works
MCP to ATDF Converter: Tools to convert from MCP format
Contributing: How to contribute to the format
Changelog: History of changes

GitHub Repository

Changelog