Repolode
Submit
agentscope-ai

agentscope-ai/agentscope

A Python agent framework for building, running and observing LLM agents, with tools, multi-agent teams, memory, sandboxed execution and a deployable agent service.

View on GitHubWebsite
Stars
35,511
Forks
3,805
Open issues
670
Last commit
3 days ago

What is it?

AgentScope is a Python framework for building LLM-based agents. Its SDK layer provides a ReAct reasoning-and-acting loop, a toolkit that accepts Python functions, MCP servers and skills, and model adapters for providers such as OpenAI, Anthropic, Gemini, DashScope, DeepSeek and Ollama. Context management, middleware hooks, permission control with human confirmation, and pluggable memory backends such as ReMe and Mem0 are included as building blocks.

For execution, agents can run tools and code in isolated sandboxes, including local, Docker, E2B and Kubernetes environments. An event system streams reasoning, tool calls and multimodal content to a frontend, and realtime voice agents are supported.

The project also ships an agent service: a FastAPI backend with a pre-built Web UI that handles multi-tenant, multi-session applications, agent teams, IM channels such as Feishu and Discord, RAG, scheduled tasks and persistence.

Key features

  • 01ReAct loop with structured output, interruption and batched tool calls
  • 02Toolkit supporting Python tools, MCP servers and skills, with built-in shell, file and search tools
  • 03Model support across OpenAI, Anthropic, Gemini, DashScope, DeepSeek, Ollama and others
  • 04Sandboxed tool and code execution in local, Docker, E2B or Kubernetes environments
  • 05Agent service with a pre-built Web UI, multi-tenant sessions and IM channel integration

Use cases

  • Building a multi-step agent that calls tools and edits files in a sandbox
  • Creating a team of leader and worker agents that split up a complex task
  • Deploying a multi-session chat application backed by agents with a ready-made web frontend
  • Prototyping a realtime voice assistant that streams responses

Quick start

  1. 1

    Install AgentScope from PyPI (requires Python 3.11+)

    uv pip install agentscope
  2. 2

    Save the console example from the README and run it with a DASHSCOPE_API_KEY set

Things to know

  • Requires Python 3.11 or higher.
  • The sample agent needs a DashScope API key from a model provider.
  • Some features such as realtime voice and SOP are marked experimental.

In collections

Multi-agent orchestration frameworksOpen-source AI agent frameworks & appsMCP serversAI chatbots & chat clients

Topics

#agent#llm#multi-agent#mcp#python#tool-use#sandbox#voice-agent#chatbot#large-language-models

FAQ

What is agentscope?
A Python agent framework for building, running and observing LLM agents, with tools, multi-agent teams, memory, sandboxed execution and a deployable agent service. AgentScope is a Python framework for building LLM-based agents. Its SDK layer provides a ReAct reasoning-and-acting loop, a toolkit that accepts Python functions, MCP servers and skills, and model adapters for providers such as OpenAI, Anthropic, Gemini, DashScope, DeepSeek and Ollama. Context management, middleware hooks, permission control with human confirmation, and pluggable memory backends such as ReMe and Mem0 are included as building blocks.
Is agentscope free to use? What license does it have?
agentscope is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is agentscope still maintained?
Actively maintained. The last commit to agentscope was on Nov 4, 2026.
What can I use agentscope for?
Building a multi-step agent that calls tools and edits files in a sandbox Creating a team of leader and worker agents that split up a complex task Deploying a multi-session chat application backed by agents with a ready-made web frontend Prototyping a realtime voice assistant that streams responses
How do I install agentscope?
Install AgentScope from PyPI (requires Python 3.11+): uv pip install agentscope Save the console example from the README and run it with a DASHSCOPE_API_KEY set
Where can I get agentscope?
The source code and installation instructions are on GitHub: https://github.com/agentscope-ai/agentscope

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects