agentscope

What it is and what it does

AgentScope is a framework for building and deploying multi-agent systems powered by large language models. It provides composable building blocks—agents with reasoning-acting loops, tool management (Python tools, MCP servers, skills), model integration across major providers, context management with automatic compaction, and an event system for streaming reasoning and tool calls. The framework is designed to work with increasingly capable agentic LLMs, leveraging their reasoning and tool-use abilities rather than constraining them with rigid orchestration.

Beyond the SDK layer, AgentScope includes a batteries-included agent service—a FastAPI backend with a pre-built Web UI—that turns agents into multi-tenant, multi-session applications. It supports agent teams with leader-worker orchestration, channels for connecting to messaging platforms (Feishu, Discord), RAG services, and a hub system for discovering and installing MCP servers and skills. The framework also provides isolated execution environments (local, Docker, K8s, E2B, Daytona) for safe tool and code execution.

Use it for:

• Build a reasoning agent that breaks down tasks, calls tools (shell, file edit, search), and streams results to a frontend.
• Deploy a multi-agent team service where specialized agents coordinate on complex tasks with leader-worker orchestration.
• Connect agents to messaging platforms (Feishu, Discord) to answer questions and execute tasks from chat channels.
• Implement RAG-augmented agents that retrieve and reason over documents in a multi-tenant service.
• Compose agents with long-term memory (ReMe, Mem0) to maintain context across sessions.
• Integrate with multiple LLM providers (OpenAI, Anthropic, DashScope, DeepSeek) without rewriting agent logic.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

AgentScope is a production-ready framework for building multi-agent systems with LLMs, providing abstractions for agent composition, tool management, model integration, and service deployment.

Yes. AgentScope is actively maintained, well-documented, and designed for production use. It offers a comprehensive toolkit for multi-agent systems with low install friction and permissive licensing. The 25 runtime dependencies are substantial but necessary for the breadth of features (LLM APIs, observability, async I/O, code parsing). Install if you are building agents that need tool use, multi-agent coordination, or service deployment; skip if you need only simple LLM chat.

Install

agentscope on PyPI

pip install agentscope

uv add agentscope

poetry add agentscope

Installing agentscope

Before you install

Low friction installation; pure Python wheel. Active maintenance with recent release (7 days old) and strong community signal (28949 GitHub stars). Requires Python 3.11 or higher.

License in practice

Apache-2.0 permissive license allows commercial use, modification, and distribution with minimal restrictions—suitable for production deployment.

Quickstart

pip install agentscope

from agentscope.agent import Agent
from agentscope.model import DashScopeChatModel
from agentscope.credential import DashScopeCredential
from agentscope.message import UserMsg

agent = Agent(
    name="Assistant",
    model=DashScopeChatModel(
        credential=DashScopeCredential(api_key="..."),
        model="qwen3.6-plus"
    )
)
await agent.reply_stream(UserMsg("user", "Hello"))

Requires Python 3.11 or higher; LLM API credentials (e.g., DashScope, OpenAI, Anthropic) needed to run agents.

Verify before relying

• Performance characteristics and latency for multi-agent orchestration at scale.
• Compatibility and integration depth with specific LLM providers beyond those documented.
• Production deployment guidance for high-concurrency multi-tenant scenarios.

Package facts

Evidence: agentscope-2.0.6-py3-none-any.whl

Keywords: agent, multi-agent, LLM, AI

Tags

More Artificial Intelligence packages

Further reading