AgentScope is a flexible and robust multi-agent platform designed to simplify the development and deployment of multi-agent applications. It provides a message exchange communication mechanism as its core, along with a range of features to lower the barriers to development and deployment. These include abundant syntactic tools, built-in agents and service functions, user-friendly interfaces for application demonstration and utility monitoring, a zero-code programming workstation, and an automatic prompt tuning mechanism. AgentScope also provides both built-in and customizable fault tolerance mechanisms, system-level support for managing and utilizing multi-modal data, tools, and external knowledge, and an actor-based distribution framework that enables easy conversion between local and distributed deployments and automatic parallel optimization. AgentScope is designed with a strong emphasis on usability, providing a procedure-oriented message exchange mechanism, a zero-code drag-and-drop programming workstation, and automatic prompt tuning mechanisms. It also offers robust fault tolerance for diverse LLMs and APIs, extensive compatibility for multi-modal, tools, and external knowledge, and optimized efficiency for distributed multi-agent operations. The platform supports multi-modal data, tools, and external knowledge, and provides a flexible solution for agents to handle different information. The actor-based distributed mode of AgentScope can help develop efficient and reliable distributed multi-agent applications seamlessly. AgentScope provides a rich set of built-in resources, including services, dedicated agents, and pre-configured examples, to reduce the initial setup effort and enable rapid prototyping and deployment of multi-agent LLM systems. It also introduces interaction interfaces tailored for multi-agent systems, providing a rich multi-modal experience, crucial for systems incorporating LLMs that handle diverse data types. The platform includes a monitoring module that tracks model and API usage, as well as calculating financial costs, and provides a Gradio-based interface for graphical application development. AgentScope also provides automatic prompt tuning, allowing users to generate prompts based on a simple description of the agent in natural language, update prompts according to contexts, and enable in-context learning. The platform includes a logging system for developers to quickly monitor and identify problems in multi-agent applications. AgentScope supports multi-modal applications, allowing agents to perceive, change their environment, and handle more complex tasks. The platform also provides a tool usage module based on the ReAct algorithm, which allows for the generation of interleaved reasoning and task-specific actions, along with a core component—service toolkit. This design features high compatibility, extensibility, robustness, and re-usability, spanning from function pre-processing, prompt engineering, reasoning, and response parsing to agent-level fault tolerance.AgentScope is a flexible and robust multi-agent platform designed to simplify the development and deployment of multi-agent applications. It provides a message exchange communication mechanism as its core, along with a range of features to lower the barriers to development and deployment. These include abundant syntactic tools, built-in agents and service functions, user-friendly interfaces for application demonstration and utility monitoring, a zero-code programming workstation, and an automatic prompt tuning mechanism. AgentScope also provides both built-in and customizable fault tolerance mechanisms, system-level support for managing and utilizing multi-modal data, tools, and external knowledge, and an actor-based distribution framework that enables easy conversion between local and distributed deployments and automatic parallel optimization. AgentScope is designed with a strong emphasis on usability, providing a procedure-oriented message exchange mechanism, a zero-code drag-and-drop programming workstation, and automatic prompt tuning mechanisms. It also offers robust fault tolerance for diverse LLMs and APIs, extensive compatibility for multi-modal, tools, and external knowledge, and optimized efficiency for distributed multi-agent operations. The platform supports multi-modal data, tools, and external knowledge, and provides a flexible solution for agents to handle different information. The actor-based distributed mode of AgentScope can help develop efficient and reliable distributed multi-agent applications seamlessly. AgentScope provides a rich set of built-in resources, including services, dedicated agents, and pre-configured examples, to reduce the initial setup effort and enable rapid prototyping and deployment of multi-agent LLM systems. It also introduces interaction interfaces tailored for multi-agent systems, providing a rich multi-modal experience, crucial for systems incorporating LLMs that handle diverse data types. The platform includes a monitoring module that tracks model and API usage, as well as calculating financial costs, and provides a Gradio-based interface for graphical application development. AgentScope also provides automatic prompt tuning, allowing users to generate prompts based on a simple description of the agent in natural language, update prompts according to contexts, and enable in-context learning. The platform includes a logging system for developers to quickly monitor and identify problems in multi-agent applications. AgentScope supports multi-modal applications, allowing agents to perceive, change their environment, and handle more complex tasks. The platform also provides a tool usage module based on the ReAct algorithm, which allows for the generation of interleaved reasoning and task-specific actions, along with a core component—service toolkit. This design features high compatibility, extensibility, robustness, and re-usability, spanning from function pre-processing, prompt engineering, reasoning, and response parsing to agent-level fault tolerance.