What is AgentScope?
AgentScope is an open-source Python framework for developers building AI agents and agent applications. AgentScope provides components for reasoning, tool use, context management, and application delivery. The framework emphasizes using language models’ reasoning and tool-use capabilities rather than constraining agents through strict prompts and prescribed orchestration.
AgentScope separates agent construction from application serving. The SDK supplies building blocks for individual agents, while the accompanying agent service adds a FastAPI backend and a pre-built Web UI. This structure supports both terminal-based agent development and applications with multiple tenants and isolated sessions.
What AgentScope can do
AgentScope combines model access, tool execution, memory, and coordination within composable agent components. AgentScope also exposes events for streaming reasoning, tool calls, and multimodal content to a frontend. Developers can extend the agent loop through middleware hooks for model calls, permissions, context compression, and other stages.
- AgentScope runs ReAct loops with structured output, interruption, and resume.
- AgentScope connects Python tools, MCP servers, and skills.
- AgentScope includes shell, file editing, search, and planning tools.
- AgentScope coordinates leader agents and worker agents.
- AgentScope supports ReMe and Mem0 memory backends.
- AgentScope provides automatic context compaction and tool-result offloading.
- AgentScope supports scheduled tasks and background task offloading.
How AgentScope works
AgentScope can be installed from PyPI using uv pip or from a cloned GitHub repository in editable mode. AgentScope requires Python 3.11 or higher. The terminal example creates an agent with a system prompt, a DashScope chat model, credentials read from an environment variable, and a toolkit of coding tools.
AgentScope provides a console launcher that handles streamed output, tool-call confirmation, and interruption. For a web application, the repository instructs developers to start the Python service backend and then run the Web UI separately with pnpm. Execution options include local environments, Docker, Apple Container, Bubblewrap, E2B, OpenSandbox, Daytona, and K8s.
Who AgentScope is for
AgentScope is intended for developers building agentic applications, including assistants that use tools and teams that delegate work between agents. AgentScope’s service capabilities also support applications needing retrieval, stored sessions, shared resources, or messaging channels. The supplied examples cover terminal interaction and a browser-based interface.
- AgentScope supports coding assistants that run shell commands and edit files.
- AgentScope supports teams that delegate tasks through leader–worker coordination.
- AgentScope supports retrieval applications with blob storage and an index worker.
- AgentScope connects messaging assistants to Feishu, Discord, or custom channels.
- AgentScope supports applications that share models, skills, and workspaces across groups.
AgentScope pricing and access
AgentScope is an open-source project released under Apache License 2.0, with access through PyPI or the GitHub source repository. The supplied repository text does not list commercial AgentScope subscription plans. The included quickstart shows package installation, source installation, and commands for starting the agent service and Web UI.
AgentScope strengths and limitations
AgentScope combines agent development primitives with application-serving capabilities, reducing the need to assemble those layers separately. AgentScope supports model providers including OpenAI, Anthropic, Gemini, DashScope, DeepSeek, Moonshot, Volcengine, xAI, and Ollama. Permission controls, confirmation prompts, and isolated execution options provide explicit mechanisms for managing tool access.
- AgentScope requires Python 3.11 or higher.
- AgentScope requires technical setup through installation, configuration, and service commands.
- AgentScope marks realtime voice features and standard operating procedure support as experimental.
AgentScope FAQ
What is AgentScope?
AgentScope is an open-source Python framework for building AI agents and multi-agent applications. AgentScope includes reasoning-and-acting loops, tool management, model integrations, context handling, memory, and an agent service with a FastAPI backend and pre-built Web UI.
Is AgentScope free?
AgentScope is available as an open-source project under Apache License 2.0. AgentScope can be installed from PyPI or GitHub source, and the supplied repository text does not list AgentScope subscription prices or commercial plans.
What are the requirements for running AgentScope?
AgentScope requires Python 3.11 or higher and supports installation from PyPI or source. AgentScope’s web application instructions start the Python backend separately from the Web UI, which uses pnpm install and pnpm dev.
Does AgentScope support MCP and model integrations?
AgentScope supports MCP servers, Python tools, and skills through its Toolkit. AgentScope also integrates models from providers including OpenAI, Anthropic, Gemini, DashScope, DeepSeek, Moonshot, Volcengine, xAI, and Ollama, with model capabilities covering language, embeddings, and text to speech.
Can AgentScope run multi-agent teams?
AgentScope supports leader–worker teams in which a leader agent delegates tasks to member agents. AgentScope also includes task planning, built-in team tools, and background task offloading, allowing a long-running tool result to wake an agent and resume the conversation.
Does AgentScope support voice agents?
AgentScope supports experimental realtime voice agents and integrations with DashScope, OpenAI, Gemini, and xAI realtime APIs. AgentScope also lists experimental agent-service voice conversations over WebRTC, with text and voice sharing the same session history.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




