AgentScope
Build and run agents you can see, understand and trust.
What is AgentScope?
AgentScope is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to build and run agents you can see, understand and trust.
Build and run agents you can see, understand and trust.
This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- Build and run agents you can see, understand and trust.
Use Cases
Maintainer
Works with
Installation
Manual Installation
npx agentscopeConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use AgentScope
AgentScope 2.0 is a production-ready Python framework for building transparent, observable multi-agent systems powered by large language models. It provides a unified event bus, a fine-grained permission system, multi-tenancy support, sandbox/workspace isolation (local, Docker, E2B), and an extensible middleware system for the agent's reasoning-acting loop. Rather than constraining agents with rigid orchestration scripts, AgentScope leverages models' native reasoning and tool-use abilities, making it suitable for building autonomous agents for complex business workflows. Developers use it to create verifiable, trustworthy agents where each decision and action is observable and auditable.
Prerequisites
- Python 3.11 or higher installed
- An LLM provider API key (e.g., DASHSCOPE_API_KEY for Qwen models, or equivalent for OpenAI/Anthropic)
- uv or pip package manager for Python
- An MCP-compatible client such as Claude Desktop or Cursor
- Docker (optional, for sandbox workspace isolation)
Install AgentScope from PyPI
Install the agentscope package using uv (recommended) or pip. AgentScope requires Python 3.11+ due to its use of modern async features and type annotation syntax.
uv pip install agentscope
# or
pip install agentscopeSet your LLM provider credentials
AgentScope supports multiple LLM providers. Set the API key for your chosen provider as an environment variable. The example uses DashScope (Alibaba Cloud) for Qwen models, but OpenAI and Anthropic are also supported.
export DASHSCOPE_API_KEY=your-dashscope-api-key
# Or for OpenAI:
# export OPENAI_API_KEY=your-openai-api-keyBuild your first agent
Create a Python script using AgentScope's Agent class with a Toolkit containing tools like Bash, Grep, Glob, Read, Write, and Edit. Assign it a model and system prompt to define its role and capabilities.
from agentscope.agent import Agent
from agentscope.tool import Toolkit, Bash, Read, Write
from agentscope.credential import DashScopeCredential
from agentscope.model import DashScopeChatModel
import os, asyncio
async def main():
agent = Agent(
name="Friday",
system_prompt="You're a helpful assistant.",
model=DashScopeChatModel(
credential=DashScopeCredential(api_key=os.environ["DASHSCOPE_API_KEY"]),
model="qwen3.6-plus",
),
toolkit=Toolkit(tools=[Bash(), Read(), Write()]),
)
asyncio.run(main())Add the AgentScope MCP server to your client configuration
Configure the AgentScope MCP server in your MCP client's configuration file. This allows your AI coding assistant to build and orchestrate AgentScope agents through the MCP interface.
Restart your MCP client and start building agents
After saving the configuration and restarting your MCP client, the AgentScope server is available. You can now ask your AI assistant to scaffold multi-agent systems, configure tool permissions, or set up agent service deployments using AgentScope patterns.
AgentScope Examples
Client configuration
Add this block to your claude_desktop_config.json under the mcpServers key to enable the AgentScope MCP server.
{
"mcpServers": {
"agentscope": {
"command": "npx",
"args": ["agentscope"]
}
}
}Prompts to try
Once connected, use these prompts with your AI assistant to leverage AgentScope's multi-agent capabilities.
- "Build an AgentScope agent team where a planner agent delegates research tasks to two worker agents"
- "Create an AgentScope agent with sandbox isolation using Docker for running untrusted code"
- "Set up an AgentScope multi-session agent service with tenant isolation for a SaaS product"
- "Add a permission system to my AgentScope agent to restrict which tools it can call"Troubleshooting AgentScope
ImportError or ModuleNotFoundError when importing agentscope
Verify you are running Python 3.11+. Run 'python --version' to check. If needed, use pyenv or conda to switch to a compatible version before installing agentscope.
Agent hangs or times out when calling the LLM
Check that your API key environment variable is set correctly and that your account has sufficient quota. For DashScope, verify the model name (e.g., 'qwen3.6-plus') is available in your region at dashscope.console.aliyun.com.
Sandbox/Docker workspace errors when running Bash tool
Ensure Docker Desktop is running and the Docker daemon is accessible. For E2B sandbox, set the E2B_API_KEY environment variable. For local execution, the Bash tool runs in the current process without sandbox by default.
Frequently Asked Questions about AgentScope
What is AgentScope?
AgentScope is a Model Context Protocol (MCP) server that build and run agents you can see, understand and trust. It connects AI assistants to external tools and data sources through a standardized interface.
How do I install AgentScope?
Follow the installation instructions on the AgentScope GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.
Which AI clients work with AgentScope?
AgentScope works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is AgentScope free to use?
Yes, AgentScope is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.
AgentScope Alternatives — Similar Coding Agents Servers
Looking for alternatives to AgentScope? Here are other popular coding agents servers you can use with Claude, Cursor, and VS Code.
Dify
★ 142.2kProduction-ready platform for agentic workflow development.
Ruflo
★ 54.0k🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, self-learning swarm intelligence, RAG integrat
Goose
★ 45.7kan open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM
Antigravity Awesome Skills
★ 38.3kInstallable GitHub library of 1,400+ agentic skills for Claude Code, Cursor, Codex CLI, Gemini CLI, Antigravity, and more. Includes installer CLI, bundles, workflows, and official/community skill collections.
Serena
★ 24.5kA coding agent toolkit that provides IDE-like semantic code retrieval and editing tools, enabling LLMs to efficiently navigate and modify codebases using symbol-level operations instead of basic file reading and string replacements.
Gemini
★ 11.6kEnables comprehensive codebase analysis using Google's Gemini CLI and its massive context window. Supports file/directory analysis, security audits, architecture analysis, feature verification, and complete project overviews for large codebases that
Browse More Coding Agents MCP Servers
Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up AgentScope in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use AgentScope?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.