Llmagentnet: A Collaborative Network of Autonomous AI Agents for Complex Task Execution

Recent advancements in Large Language Models (LLMs) have opened new horizons for building autonomous AI agents capable of executing complex, multi-step tasks across a wide range of domains. Despite this progress, existing multi-agent systems often suffer from limited interoperability and extensibility due to rigid communication protocols, closed ecosystems, and monolithic architectural assumptions. To address these limitations, we introduce LLMAgentNet – a collaborative, modular framework designed for orchestrating multiple autonomous AI agents that leverage LLMs for planning, reasoning, and execution. The LLMAgentNet framework is based on a decentralized agent architecture in which agents can communicate and collaborate using standardized RESTful API calls. By utilizing the OpenAPI Specification, LLMAgentNet enables seamless integration of both native and external agents, including those developed using other frameworks or hosted on different platforms. Each agent in the system is capable of interpreting high-level user requests, breaking them down into sub-tasks, selecting appropriate tools or agents for execution, and maintaining contextual memory across task sequences. A distinctive feature of LLMAgentNet is its hierarchical Agent Manager (Orchestrator), which coordinates interactions, maintains execution trajectories, and optimizes the use of specialized agents. The system supports multiple operational modes – Human-in-the-Loop, Human-on-the-Loop, and Human-out-of-the-Loop – enabling both supervised development and fully autonomous operation. Notably, LLMAgentNet includes a memory mechanism that allows agents to access and update knowledge bases via vector databases (e.g., ChromaDB), supporting long-term memory, runtime variables, and contextual reasoning. The framework has been benchmarked using standard datasets such as HotPotQA and in practical use cases such as content marketing automation, demonstrating significant improvements in efficiency and quality over manual or single-agent approaches. By enabling collaboration, reflection, and self-improvement among AI agents, LLMAgentNet contributes to the development of general-purpose intelligent systems and marks a critical step toward the realization of Artificial General Intelligence (AGI).