Python Agents MCP Course

具有模型上下文协议的AI代理-Python代码示例

此存储库包含Coursera课程“具有模型上下文协议的AI代理”的Python代码示例

先决条件

• Python 3.10+
• 点
• 来自OpenAI、Anthropic或Google的API密钥（设置于 .env 文件）

设置

每个模块都有自己的 python-code/ 目录。要设置任何模块，请执行以下操作：

cd /python-code
pip install -r requirements.txt
cp .env.example .env  # Then add your API key(s)

______________________________________________________________________

课程结构和代码映射

模块1：人工智能代理模型上下文协议（MCP）入门

*模块1是概念介绍，没有代码练习。*

______________________________________________________________________

模块2:AI代理循环和模型上下文协议

代码位置： 01-intro-to-mcp-agents/python-code/

关键文件：

命令：

cd 01-intro-to-mcp-agents/python-code
pip install -r requirements.txt
python server.py              # Start the basic MCP server
python agent.py "question"    # Run the basic agent
python test_server.py         # Test server tools directly

______________________________________________________________________

模块3：使用模型上下文协议构建AI代理

代码位置： 01-intro-to-mcp-agents/python-code/

关键文件：

命令：

cd 01-intro-to-mcp-agents/python-code
python workspace_server.py           # Start workspace-aware server
python workspace_agent.py "question" # Run workspace agent with context discovery

______________________________________________________________________

模块4:AI代理的稳健错误处理技术

代码位置： 02-failing-forward/python-code/

关键文件：

命令：

cd 02-failing-forward/python-code
pip install -r requirements.txt
python expense_server.py           # Start the expense server
python expense_agent.py "request"  # Run the expense agent
python test_failing_forward.py     # Run failing forward tests

模式映射：

______________________________________________________________________

模块5：更快、更可预测、更有能力的人工智能代理

代码位置： 03-intelligence-budget/python-code/

关键文件：

命令：

cd 03-intelligence-budget/python-code
pip install -r requirements.txt
python agent_heavy_server.py   # Start agent-heavy server
python tool_heavy_server.py    # Start tool-heavy server
python hybrid_server.py        # Start hybrid server
python test_all.py             # Run comparison tests

模式映射：

______________________________________________________________________

快速参考：Coursera项目→ 代码文件

______________________________________________________________________

环境设置

创建一个 .env 每个模块中的文件 python-code/ 目录：

OPENAI_API_KEY=your-openai-key-here
ANTHROPIC_API_KEY=your-anthropic-key-here
GOOGLE_API_KEY=your-google-key-here

您只需要为计划使用的LLM提供程序提供API密钥。

______________________________________________________________________

项目结构

code-repo/
├── README.md
├── .gitignore
│
├── 01-intro-to-mcp-agents/
│   └── python-code/
│       ├── server.py              # Module 2: Building Your First MCP Server
│       ├── agent.py               # Module 2: Building Your First MCP AI Agent
│       ├── llm.py                 # LLM integration utilities
│       ├── workspace_server.py    # Module 3: All workspace/context items
│       ├── workspace_agent.py     # Module 3: Workspace-related Guidance
│       ├── server_with_resources.py  # Server with MCP resources
│       ├── agent_with_learning.py    # Agent with active learning
│       ├── test_server.py         # Server testing
│       ├── test_resources.py      # Resource testing
│       ├── requirements.txt       # Python dependencies
│       ├── README.md              # Setup instructions
│       └── workspace/             # Sample workspace with .context.md files
│
├── 02-failing-forward/
│   └── python-code/
│       ├── expense_server.py      # Module 4: All error handling items
│       ├── expense_agent.py       # Agent for expense workflows
│       ├── test_failing_forward.py # Pattern tests
│       ├── requirements.txt       # Python dependencies
│       └── README.md              # Setup instructions
│
└── 03-intelligence-budget/
    └── python-code/
        ├── agent_heavy_server.py           # Module 5: Managing Cognitive Load
        ├── tool_heavy_server.py            # Module 5: Managing Cognitive Load
        ├── hybrid_server.py                # Balanced approach
        ├── scripted_orchestration_server.py # Module 5: Scripted Orchestration
        ├── self_prompting_server.py        # Module 5: Self-Prompting
        ├── validate_at_source_server.py    # Module 5: Tool Design for Errors
        ├── database.py                     # Shared mock database
        ├── test_all.py                     # Comparison tests
        ├── requirements.txt                # Python dependencies
        └── README.md                       # Setup instructions

______________________________________________________________________

许可证

麻省理工学院