ToolRegistry: A Protocol-Agnostic Tool Management Library for Function-Calling LLMs¶

Welcome to the official documentation for ToolRegistry, a Python library designed to register, manage, and execute tools (functions) for Agentic AI and large language model applications. This documentation is organized by topic to help you quickly find and learn about different aspects of the library.

Overview¶

ToolRegistry is a powerful Python library that simplifies the registration, management, and execution of tools (functions). Whether you're building systems integrated with large language models or need structured tool management, ToolRegistry provides a consistent interface supporting both synchronous and asynchronous tool calls.

Key Features¶

• Protocol-Agnostic: Works with OpenAI, MCP, OpenAPI, LangChain, and more
• Multi-Provider Schema: Generate tool schemas for OpenAI, Anthropic, and Gemini via llm-rosetta
• Unified Interface: Consistent API for sync and async tool execution
• Permission System: Rule-based authorization with composable policies, built-in tag matching, and handler protocols for interactive approval
• Enable/Disable Control: Fine-grained tool availability management with reason tracking
• Tool Metadata & Tags: Classify tools with predefined and custom tags for policy-driven control
• Namespace Support: Organize tools into logical groups with MRO inheritance
• Admin Panel: Built-in web-based administration interface with execution logging, REST API, and Anthropic-style minimalist UI
• Callback Mechanism: Monitor registry changes with event-driven callbacks

This documentation details how to install, configure, and use the library in real projects. Browse the sections in the left menu to find the information you need.

Getting Started¶

Quickly start using ToolRegistry by installing it with this command:

pip install toolregistry

Detailed setup instructions →

Ecosystem¶

ToolRegistry is part of a three-package ecosystem. See the Ecosystem page for details.

Documentation Contents¶

Here are brief introductions and links to each section:

• Installation Guide Detailed instructions for installing ToolRegistry, including basic installation and MCP-supported installation.

• Basic Usage Guide Provides getting started examples and tutorials to help you quickly learn tool registration, invocation, and management.

• Function calling via OpenAI Compatible API Explains how to use ToolRegistry with OpenAI compatible API through function calling.

• Register Tools from Various Sources

• MCP Tool Integration

• OpenAPI Tool Integration

• Class-based Tools Collection

• Hub Tools Collection

• LangChain Tool Integration

• Permission System Control tool execution with composable rules, tag-based policies, and handler protocols.

• Concurrency Modes Explains thread and process execution modes and their performance characteristics.

• Best Practices on Tool Implementation Offers principles and recommendations for designing and implementing tools.

• Examples Demonstrates practical use cases, including consecutive tool call examples.

• API References Comprehensive API documentation for all classes and methods in ToolRegistry.

Citation¶

If you use ToolRegistry in your research or project, please consider cite it as:

@software{toolregistry2025,
  title={ToolRegistry: A Protocol-Agnostic Tool Management Library for OpenAI-Compatible LLM Applications},
  author={Peng Ding},
  year={2025},
  url={https://github.com/Oaklight/ToolRegistry},
  note={A Python library for unified tool registration, execution, and management across multiple protocols in OpenAI-compatible LLM applications}
}

@article{ding2025toolregistry,
  title={ToolRegistry: A Protocol-Agnostic Tool Management Library for Function-Calling LLMs},
  author={Ding, Peng},
  journal={arXiv preprint arXiv:2507.10593},
  year={2025}
}

License¶

ToolRegistry is licensed under the MIT License.