Tablestore MCP servers.
- 入门示例: tablestore-java-mcp-server
- 基于 MCP 架构实现知识库答疑系统: tablestore-java-mcp-server-rag
- 实现一个目前最常见的一类 AI 应用即答疑系统，支持基于私有知识库的问答，会对知识库构建和 RAG 做一些优化。
欢迎加入我们的钉钉公开群，与我们一起探讨 AI 技术。钉钉群号：36165029092
MCP.Pizza Chef: aliyun
Four sample projects from Alibaba Cloud share this repository, two in Java and two in Python. The ready-made one is the Python memory server, published for anyone to install, and it does exactly two things: save a note, and find earlier notes that mean something similar. Running it needs a paid Alibaba Cloud Tablestore instance along with its access key and secret, and most of the documentation is written in Chinese. By default it listens on every network address with no password.
Data
Notes
Use This MCP server To
Save notes my assistant can recall in a later chat Find past notes by meaning rather than exact words Give my assistant memory that survives between sessions Store and search a Chinese-language knowledge base Keep answers to common questions in one searchable place
README
alibabacloud-tablestore-mcp-server FAQ
Can I use this to give my assistant a memory?
Yes. It saves whatever you ask it to remember and finds it again later by meaning rather than by exact wording.
What does it cost?
The code is free, but it stores everything in Alibaba Cloud Tablestore, a paid cloud service with only a limited free allowance.
What credential does it need?
Not a simple key. It wants an Alibaba Cloud access key and secret pair, plus the name and address of the Tablestore instance you created.
Is it safe to run exactly as it comes?
Not on a shared network. It listens on every network address on port 8001 with no password, so anyone who can reach that port can read and write your notes. Start it with the stdio option instead.
How many tools does it actually give me?
Two in the Python server, one to save and one to search. The other projects in the repository are worked examples rather than finished products.
Is the documentation in English?
Mostly not. The front page and the built-in tool descriptions are written in Chinese, so expect to translate as you go.
How hard is setup?
Hard. You create a cloud database, set roughly half a dozen configuration values, and on first run it downloads a large text-matching model onto your own machine.