https://github.com/abhimvp/CustomMCPServer
build a custom MCP server in Python and connect that to an AI agent.
https://github.com/abhimvp/CustomMCPServer
Last synced: about 2 months ago
JSON representation
build a custom MCP server in Python and connect that to an AI agent.
- Host: GitHub
- URL: https://github.com/abhimvp/CustomMCPServer
- Owner: abhimvp
- Created: 2025-04-16T14:38:09.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-04-16T16:20:51.000Z (7 months ago)
- Last Synced: 2025-04-17T00:21:03.479Z (7 months ago)
- Language: Python
- Size: 338 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- awesome-mcp-servers - **CustomMCPServer** - build a custom MCP server in Python and connect that to an AI agent. `python` `mcp` `server` `ai` `pip install git+https://github.com/abhimvp/CustomMCPServer` (🤖 AI/ML)
- awesome-mcp-servers - **CustomMCPServer** - build a custom MCP server in Python and connect that to an AI agent. `python` `mcp` `server` `ai` `pip install git+https://github.com/abhimvp/CustomMCPServer` (AI/ML)
README
# CustomMCPServer
Build a custom MCP server in Python and connect that to an AI agent.
- Resources:
- [Python MCP Server SDK](https://github.com/modelcontextprotocol/python-sdk?tab=readme-ov-file)
- [Reference Code](https://github.com/techwithtim/PythonMCPServer)
- MCP servers are essentially how LLM and AI Tools communicate with each other.The Model Context Protocol allows applications to provide context for LLMs in a standardized way, separating the concerns of providing context from the actual LLM interaction.
```bash
# create a uv-managed project
$ uv init .
Initialized project `custommcpserver` at `C:\Users\abhis\Desktop\AIAgents\CustomMCPServer`
# Then add MCP to your project dependencies:
$ uv add "mcp[cli]"
# activate python environment as well
$ source .venv/scripts/activate
# You can install this server in Claude Desktop and interact with it right away by running:
# mcp install main.py
# since we're in this vs code we do
$ uv run mcp install main.py
[04/16/25 10:06:31] INFO Added server 'Demo' to Claude config
INFO Successfully installed Demo in Claude app
# go to claude app > File > settings > Developer > Edit Config - opens up some file and open that claude_desktop_config.json file
# we can see the mcp run main.py(path) is added in that file.
# the mcp server should show up on claude app - if it's not - we do the following
# close claude app - open task manager and search fo claude and right click & do end task.
# wait a sec & open the app again
# once the app loads , it spins up the mcp server and it shows on app - refer the image below & when you click on tool we can see the
# add tool and server:DEMO
```
## Advanced MCP Server
- before we go ahead , **fundamentally What is MCP?**:
- lets you build **servers** that _expose data_ and functionality to LLM applications in a secure, standardized way. Think of it like a web API, but specifically designed for LLM interactions.
- Expose data through **Resources** (think of these sort of like GET endpoints; they are used to load information into the LLM's context)
- Provide functionality through **Tools** (sort of like POST endpoints; they are used to execute code or otherwise produce a side effect)
- Define interaction patterns through **Prompts** (reusable templates for LLM interactions)
- And more!
```bash
uv add black
# after creating new mcp server & tool
$ uv run mcp install adv_main.py
[04/16/25 10:44:02] INFO Added server 'AI Sticky Notes' to Claude config
INFO Successfully installed AI Sticky Notes in Claude app
```
- from this above example , we can see how to make use of tool we create and do whatever we can like.
- add a tool to connect to db and ask claude to create 100 fake users into db , instead of us manually copy paste ..etc.we can automate lot of stuff.