Internet of Agents (IoA) is an open-source framework developed by researchers from Tsinghua University, Peking University, Beijing University of Posts and Telecommunications, and Tencent that enables distributed AI agents to collaborate through internet-like connectivity. Inspired by how humans interact online, IoA creates a comprehensive ecosystem where diverse AI agents can connect, communicate, and work together to solve complex problems that might be challenging for individual agents to tackle alone.
Key Features
- Internet-Inspired Architecture: The framework establishes a platform where AI agents can interact and collaborate similarly to how humans connect through the internet
- Autonomous Nested Team Formation: AI agents can form teams and sub-teams independently, adapting their organization to meet the requirements of complex tasks
- Heterogeneous Agent Integration: IoA brings together agents with different skills and specializations, creating diverse collaborative networks with complementary capabilities
- Asynchronous Task Execution: Agents can work on multiple objectives simultaneously, significantly enhancing system efficiency and productivity
- Adaptive Conversation Flow: The platform employs a finite-state machine inspired by Speech Act Theory to manage and control communication between agents
- Scalable and Extensible Design: The framework easily incorporates new agent types or tackles different kinds of tasks as requirements evolve
How IoA Works
The framework functions as a comprehensive collaboration ecosystem with several core processes:
- Agent Registration and Discovery: Agents register their capabilities and can discover other agents with complementary skills within the network
- Dynamic Team Formation: The system autonomously forms teams of agents based on specific task requirements and agent capabilities
- Task Assignment and Execution: IoA manages the distribution of work among agents and coordinates their execution efforts
- Structured Communication: A comprehensive message protocol ensures efficient interaction between agents, facilitating clear information exchange
- Cross-Platform Integration: The framework integrates agents across multiple platforms, enhancing its versatility for various applications
Applications and Performance
IoA has demonstrated exceptional effectiveness across multiple domains:
- Research Collaboration: Agents collaborate to draft research papers by forming teams, assigning specific tasks, and integrating diverse contributions
- Complex Problem Solving: The framework excels in handling open-ended instruction tasks that require diverse skills and approaches
- Embodied AI Challenges: IoA shows superior performance in scenarios involving agents with different observation and action capabilities
- Enhanced Decision Making: The collaborative nature of the system leads to more robust decision-making processes through diverse agent input
Experiments indicate that IoA consistently outperforms state-of-the-art baselines in general tasks, embodied AI, and retrieval-augmented generation benchmarks.
Development Information
IoA is available as an open-source project under the Apache License 2.0, providing the infrastructure necessary for creating networked agent systems. The framework supports distributed deployment of AI agents with collaborative capabilities, enabling the creation of agent networks that can address complex problems through coordinated effort.
This framework is particularly valuable for entrepreneurs and small business owners looking to implement advanced AI solutions without extensive technical expertise. By leveraging IoA’s collaborative agent architecture, organizations can build sophisticated AI systems that tackle complex business challenges through coordinated agent teamwork.
Agent URL: https://github.com/OpenBMB/IoA