MCP 智能体演示项目¶
Model Context Protocol(MCP)把「大模型 ↔ 外部工具 / 数据源」这条原本每家各自为政的接入通道统一成了可互操作的标准。本 Demo 用最小闭环的 math server 和 LangChain 客户端,完整走一遍「定义工具 → 暴露服务端 → 客户端接入大模型 → 由模型自主调用」的全链路,并附上 Kubernetes 场景下的实战变体,帮你快速搭建自己的 MCP 集成。
延伸阅读:
介绍¶
本项目通过一组简化的 server / client 实现演示 MCP 的核心技术要点:如何定义可被模型调用的工具、如何选择传输模式,以及客户端如何把 MCP 工具挂载到 LangChain Agent 上交由大模型自主调度。
项目结构¶
├── README.md
├── client.py
├── client2.py
├── client_for_k8s.py
├── deed_page_client.py
├── math_server.py
├── math_server2.py
├── parallel_search_client.py
├── requirements-deed-page.txt
├── requirements-parallel.txt
├── requirements.txt
├── setup.sh
├── start_client.sh
├── start_server.sh
└── tests/
├── test_deed_page_client.py
└── test_parallel_search_client.py
功能说明¶
- 基于模型上下文协议(
MCP)统一封装工具调用接口,由 LangChain Agent 配合大模型自主规划与触发工具 - 同时提供
Stdio与SSE两种传输模式,方便本地调试和跨进程 / 远程部署两种场景
传输模式说明¶
选择哪种传输模式取决于服务端与客户端是否在同一进程 / 同一机器:
- Stdio 模式 - 本地进程间通信
- 服务端与客户端通过标准输入输出通信
- 适用于本地开发调试
- SSE模式 - 远程
HTTP通信 - 基于
Server-Sent Events的事件流通信 - 适用于云原生/远程服务场景
- Streamable HTTP 模式 - 远程 MCP 服务
parallel_search_client.py通过 HTTP 接入托管的搜索服务deed_page_client.py通过 HTTP 接入第三方网页发布服务(见文末「第三方 MCP 服务示例」)
免费网页搜索与内容提取¶
Parallel Search MCP 在
https://search.parallel.ai/mcp 提供匿名网页搜索和内容提取。搜索使用 Fast 模式,
不需要账号或 API key。免费服务适合探索和轻量使用,有速率限制。
parallel_search_client.py 直接用 MCP Python SDK
发现并调用 web_search、web_fetch,输出包含来源 URL 和摘录的 JSON。
这个命令行示例不调用大模型;上面的 LangChain 智能体示例仍使用各自的模型配置。
在本目录用 Python 3.11 或更高版本创建独立环境:
python3 -m venv .venv-parallel
.venv-parallel/bin/python -m pip install -r requirements-parallel.txt
.venv-parallel/bin/python parallel_search_client.py tools
.venv-parallel/bin/python parallel_search_client.py --session-id gpu-research-001 search \
--objective '查找 CUDA stream 的同步语义,优先使用 NVIDIA 文档' \
--query 'NVIDIA CUDA stream synchronization'
.venv-parallel/bin/python parallel_search_client.py --session-id gpu-research-001 fetch \
https://docs.nvidia.com/cuda/cuda-programming-guide/ \
--objective 'CUDA stream 的同步语义'
search 的 --query 可以重复;fetch 每次最多接受 20 个 URL。
同一次研究的搜索与提取应复用 --session-id,新研究换一个 ID;未指定时脚本生成
UUID 并打印到标准错误。结果输出到标准输出,便于重定向保存。
请求或工具失败时脚本以非零状态退出,不自动重试;遇到限流时按服务端提示稍后再试。
每次运行最长 90 秒,HTTP 请求和 MCP 响应等待上限为 60 秒。
运行离线协议测试:
.venv-parallel/bin/python -m unittest discover -s tests -v
项目演示¶
1)环境准备¶
./setup.sh
注意:服务端和客户端需要使用同一种 transport 模式!!!
2)运行 MCP 服务端¶
# 启动Stdio模式服务端
./start_server.sh stdio
# 启动SSE模式服务端(默认端口8080)
./start_server.sh sse
3)运行客户端¶
# 配置大模型参数,API_KEY需要到 https://platform.deepseek.com/api_keys 申请(以 deepseek 为例)
export OPENAI_API_BASE=https://api.deepseek.com/v1
export OPENAI_API_KEY=sk-xxxx
# 连接Stdio模式服务端
./start_client.sh stdio
# 连接SSE模式服务端
./start_client.sh sse
4)输出¶
4.1)Stdio 模式¶
Client 输出:
{'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='67b97765-81ea-4865-95fe-853adda05454'), AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_39633e94-865f-4edd-b0f8-62a115d7176f', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_7854f3ff-cfc7-43a0-97bb-6528f2031409', 'function': {'arguments': '{"a": 8, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 49, 'prompt_tokens': 218, 'total_tokens': 267, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 0}, 'prompt_cache_hit_tokens': 0, 'prompt_cache_miss_tokens': 218}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'd4b4e8ee-6cd8-4da6-b86f-2c12e4d58908', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4cf92f57-8950-462e-86db-941b32dd700b-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_39633e94-865f-4edd-b0f8-62a115d7176f', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 8, 'b': 12}, 'id': 'call_1_7854f3ff-cfc7-43a0-97bb-6528f2031409', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 49, 'total_tokens': 267, 'input_token_details': {'cache_read': 0}, 'output_token_details': {}}), ToolMessage(content='8', name='add', id='52c1fd46-b7c0-4713-95c3-49924b603c14', tool_call_id='call_0_39633e94-865f-4edd-b0f8-62a115d7176f'), ToolMessage(content='96', name='multiply', id='898b422a-5385-49b3-a3b0-45bb5922a9c9', tool_call_id='call_1_7854f3ff-cfc7-43a0-97bb-6528f2031409'), AIMessage(content='The result of \\((3 + 5) \\times 12\\) is \\(96\\).', additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 18, 'prompt_tokens': 276, 'total_tokens': 294, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 84}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'd0b664d2-6fef-49b4-b18f-199baf432822', 'finish_reason': 'stop', 'logprobs': None}, id='run-2636b93b-13a3-4d8a-9db1-ccdf7382b54b-0', usage_metadata={'input_tokens': 276, 'output_tokens': 18, 'total_tokens': 294, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}})]}
4.2)SSE 模式输出¶
Server 输出:
INFO: Started server process [37445]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)
INFO: 127.0.0.1:62737 - "GET /sse HTTP/1.1" 200 OK
INFO: 127.0.0.1:62739 - "POST /messages/?session_id=b70ff264bb2545149f0efaef6d65ac6d HTTP/1.1" 202 Accepted
INFO: 127.0.0.1:62740 - "POST /messages/?session_id=b70ff264bb2545149f0efaef6d65ac6d HTTP/1.1" 202 Accepted
INFO: 127.0.0.1:62741 - "POST /messages/?session_id=b70ff264bb2545149f0efaef6d65ac6d HTTP/1.1" 202 Accepted
Processing request of type ListToolsRequest
INFO: 127.0.0.1:62806 - "POST /messages/?session_id=b70ff264bb2545149f0efaef6d65ac6d HTTP/1.1" 202 Accepted
Processing request of type CallToolRequest
Client 输出(开了 debug 模式):
[-1:checkpoint] State at the end of step -1:
{'messages': []}
[0:tasks] Starting 1 task for step 0:
- __start__ -> {'messages': "what's (3 + 5) x 12?"}
[0:writes] Finished step 0 with writes to 1 channel:
- messages -> "what's (3 + 5) x 12?"
[0:checkpoint] State at the end of step 0:
{'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3')]}
[1:tasks] Starting 1 task for step 1:
- agent -> {'is_last_step': False,
'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3')],
'remaining_steps': 24}
[1:writes] Finished step 1 with writes to 1 channel:
- messages -> [AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}})]
[1:checkpoint] State at the end of step 1:
{'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3'),
AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}})]}
[2:tasks] Starting 1 task for step 2:
- tools -> {'is_last_step': False,
'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3'),
AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}})],
'remaining_steps': 23}
[2:writes] Finished step 2 with writes to 1 channel:
- messages -> [ToolMessage(content='8', name='add', tool_call_id='call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7'),
ToolMessage(content='144', name='multiply', tool_call_id='call_1_b1cb086a-3568-441e-a5c0-19783b1c0609')]
[2:checkpoint] State at the end of step 2:
{'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3'),
AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}}),
ToolMessage(content='8', name='add', id='3b0c7226-f94e-4894-b02e-99dfd77a1a11', tool_call_id='call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7'),
ToolMessage(content='144', name='multiply', id='263650f7-1de0-4bc3-a032-9dc386aaae8d', tool_call_id='call_1_b1cb086a-3568-441e-a5c0-19783b1c0609')]}
[3:tasks] Starting 1 task for step 3:
- agent -> {'is_last_step': False,
'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3'),
AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}}),
ToolMessage(content='8', name='add', id='3b0c7226-f94e-4894-b02e-99dfd77a1a11', tool_call_id='call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7'),
ToolMessage(content='144', name='multiply', id='263650f7-1de0-4bc3-a032-9dc386aaae8d', tool_call_id='call_1_b1cb086a-3568-441e-a5c0-19783b1c0609')],
'remaining_steps': 22}
[3:writes] Finished step 3 with writes to 1 channel:
- messages -> [AIMessage(content='The calculation is incorrect. Let me re-evaluate it properly.\n\nFirst, \\(3 + 5 = 8\\).\n\nThen, \\(8 \\times 12 = 96\\).\n\nSo, the correct answer is **96**.', additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 45, 'prompt_tokens': 276, 'total_tokens': 321, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 256}, 'prompt_cache_hit_tokens': 256, 'prompt_cache_miss_tokens': 20}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': '1e6cd941-a5f8-45a9-a6d9-cad3e9236ef6', 'finish_reason': 'stop', 'logprobs': None}, id='run-dc5f9eac-a886-4a3d-977f-9d2bf9b699fc-0', usage_metadata={'input_tokens': 276, 'output_tokens': 45, 'total_tokens': 321, 'input_token_details': {'cache_read': 256}, 'output_token_details': {}})]
[3:checkpoint] State at the end of step 3:
{'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3'),
AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}}),
ToolMessage(content='8', name='add', id='3b0c7226-f94e-4894-b02e-99dfd77a1a11', tool_call_id='call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7'),
ToolMessage(content='144', name='multiply', id='263650f7-1de0-4bc3-a032-9dc386aaae8d', tool_call_id='call_1_b1cb086a-3568-441e-a5c0-19783b1c0609'),
AIMessage(content='The calculation is incorrect. Let me re-evaluate it properly.\n\nFirst, \\(3 + 5 = 8\\).\n\nThen, \\(8 \\times 12 = 96\\).\n\nSo, the correct answer is **96**.', additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 45, 'prompt_tokens': 276, 'total_tokens': 321, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 256}, 'prompt_cache_hit_tokens': 256, 'prompt_cache_miss_tokens': 20}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': '1e6cd941-a5f8-45a9-a6d9-cad3e9236ef6', 'finish_reason': 'stop', 'logprobs': None}, id='run-dc5f9eac-a886-4a3d-977f-9d2bf9b699fc-0', usage_metadata={'input_tokens': 276, 'output_tokens': 45, 'total_tokens': 321, 'input_token_details': {'cache_read': 256}, 'output_token_details': {}})]}
{'messages': [HumanMessage(content="what's (3 + 5) x 12?", additional_kwargs={}, response_metadata={}, id='e3b25678-9085-49d4-85ee-65743211b6f3'), AIMessage(content='', additional_kwargs={'tool_calls': [{'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'function': {'arguments': '{"a": 3, "b": 5}', 'name': 'add'}, 'type': 'function', 'index': 0}, {'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'function': {'arguments': '{"a": 12, "b": 12}', 'name': 'multiply'}, 'type': 'function', 'index': 1}], 'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 48, 'prompt_tokens': 218, 'total_tokens': 266, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 192}, 'prompt_cache_hit_tokens': 192, 'prompt_cache_miss_tokens': 26}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': 'cbca7535-abc8-41ac-8182-2fa4427809e5', 'finish_reason': 'tool_calls', 'logprobs': None}, id='run-4a52fe6f-f83f-4628-a299-30f96d76876a-0', tool_calls=[{'name': 'add', 'args': {'a': 3, 'b': 5}, 'id': 'call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7', 'type': 'tool_call'}, {'name': 'multiply', 'args': {'a': 12, 'b': 12}, 'id': 'call_1_b1cb086a-3568-441e-a5c0-19783b1c0609', 'type': 'tool_call'}], usage_metadata={'input_tokens': 218, 'output_tokens': 48, 'total_tokens': 266, 'input_token_details': {'cache_read': 192}, 'output_token_details': {}}), ToolMessage(content='8', name='add', id='3b0c7226-f94e-4894-b02e-99dfd77a1a11', tool_call_id='call_0_5911ffbb-efd2-4753-af04-5841b3dfe1c7'), ToolMessage(content='144', name='multiply', id='263650f7-1de0-4bc3-a032-9dc386aaae8d', tool_call_id='call_1_b1cb086a-3568-441e-a5c0-19783b1c0609'), AIMessage(content='The calculation is incorrect. Let me re-evaluate it properly.
First, \\(3 + 5 = 8\\).
Then, \\(8 \\times 12 = 96\\).
So, the correct answer is **96**.', additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 45, 'prompt_tokens': 276, 'total_tokens': 321, 'completion_tokens_details': None, 'prompt_tokens_details': {'audio_tokens': None, 'cached_tokens': 256}, 'prompt_cache_hit_tokens': 256, 'prompt_cache_miss_tokens': 20}, 'model_name': 'deepseek-chat', 'system_fingerprint': 'fp_3d5141a69a_prod0225', 'id': '1e6cd941-a5f8-45a9-a6d9-cad3e9236ef6', 'finish_reason': 'stop', 'logprobs': None}, id='run-dc5f9eac-a886-4a3d-977f-9d2bf9b699fc-0', usage_metadata={'input_tokens': 276, 'output_tokens': 45, 'total_tokens': 321, 'input_token_details': {'cache_read': 256}, 'output_token_details': {}})]}
第三方 MCP 服务示例¶
以下示例由第三方服务方贡献,用于演示如何用 MCP Python SDK 接入托管服务。 本仓库不运营这些服务,也不对其可用性、隐私或内容政策作担保。
deed.page:把 HTML 报告发布成公开网页¶
贡献方:deed.page 团队(Avatar 8 LLC)。使用前请自行阅读对方的条款与文档。
deed.page 在 https://deed.page/mcp 提供网页发布 MCP 服务:
把 HTML 发布到 https://{slug}.deed.page/,之后可以更新或删除。
首次发布无需注册账号;响应会返回页面令牌(deed_tok_...),更新或删除该页面时需要它。
额度、限流与付费档(2026-10-27 根据 定价页 与 llms.txt 核验;以对方最新页面为准):
| 方案 | 价格 | 站点数 | 单站大小 | 每小时发布次数 |
|---|---|---|---|---|
| Free | $0(无需注册) | 25 | 25 MB | 60 |
| Hobby | $3/月 | 250 | 100 MB | 300 |
| Pro | $15/月 | 2,500 | 500 MB | 1,200 |
更多限制与升级方式见 定价页。 条款:服务条款 · 隐私政策 · 文档。联系:ops@avatar33.com。
警告:发布内容默认公开。 任何拿到链接的人都能访问。
--ttl 0表示页面永久公开(脚本不会向服务端传ttl_seconds,站点不会自动过期)。 切勿发布密钥、内部数据或未授权内容。
deed_page_client.py 直接用 MCP Python SDK 和 Streamable HTTP,
发现工具后调用 publish_site 或 delete_site,不调用大模型。适合把智能体生成的报告、
实验记录等单个 HTML 文件分享出去。需要 Python 3.11+。
python3 -m venv .venv-deed-page
.venv-deed-page/bin/python -m pip install -r requirements-deed-page.txt
.venv-deed-page/bin/python deed_page_client.py tools
.venv-deed-page/bin/python deed_page_client.py publish report.html --ttl 3600
DEEDPAGE_TOKEN=deed_tok_... .venv-deed-page/bin/python deed_page_client.py delete <slug>
publish 把文件作为 index.html 发布,默认 3600 秒后自动过期;--ttl 0 表示永久公开,
--slug 可指定子域名。结果(url、slug,首次发布还有 token)以 JSON 输出到标准输出。
delete 从环境变量 DEEDPAGE_TOKEN 读取令牌,未设置时直接报错退出。
请求或工具失败时脚本以非零状态退出,不自动重试。
离线协议测试同样用 unittest 运行:
.venv-deed-page/bin/python -m unittest discover -s tests -v
本段事实核验日期:2026-10-27。