MCP Servers MCP server
Commit Nov 4, 2026A collection of training resources and implementation examples for building, deploying, and consuming Model Context Protocol servers.
Also known as mpolinowski/mcp-servers
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 4, 2026
- License
- MIT
What is MCP Servers MCP?
MCP Servers is a collection of training resources and implementation examples for building, deploying, and consuming Model Context Protocol servers.
Summary generated from the project's README
About MCP Servers MCP server
Learning to build, deploy, and consume Model Context Protocol servers is the primary goal of this collection of training resources. It serves as a hands-on repository for developers who want to understand the full stack of MCP, from low-level JSON-RPC dispatchers to high-level agent integrations. The materials cover various implementation methods, including using the official Python SDK, the FastMCP framework, and TypeScript with Zod.
Users can explore different deployment strategies such as packaging servers as installable Python libraries via uv, containerizing them with Docker, or aggregating multiple upstreams behind a single endpoint using a gateway pattern. The repository provides concrete examples of servers that expose tools for file access, shell commands, and weather alerts. It also demonstrates how to consume these tools using LangChain adapters, Pydantic AI agents, and secured REST APIs featuring JWT authentication.
What you can do with it
- 1Implement TypeScript MCP serversYou can learn to write a minimal MCP server using TypeScript and the official client and server libraries.
- 2Deploy servers via DockerDevelopers can learn how to ship an MCP server as a containerized image.
- 3Create an MCP gatewayYou can build a single endpoint that proxies multiple upstream servers like DuckDuckGo.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablempolinowski/mcp-servers is public and reachablegithub.com/mpolinowski/mcp-servers ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-04github.com/mpolinowski/mcp-servers ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubModel Context Protocol (FastMCP & LangChain)
🎓 Personal training repo — hands-on notes on building & consuming MCP servers. Pre-work for the instar-mcp-server.
- MCP Server Foundation
- MCP Protocol Deepdive
- MCP Server Deployment
- MCP Gateway
- INSTAR MQTT Camera AI Agent Interface
🛠️ What is MCP used for?
MCP (Model Context Protocol) is an open, JSON-RPC-based standard for how LLM-powered apps talk to the outside world. Instead of wiring a bespoke integration per tool or data source, an MCP server exposes a set of tools (callable actions), resources (readable data), and prompts — and any MCP client (Claude, Cursor, a LangChain agent, …) can use them. One standard plug: model in, capability out.
| Layer | What it is |
|---|---|
| MCP server 🖥️ | Exposes tools / resources / prompts over stdio or streamable-HTTP. |
| MCP client 🖱️ | Connects to a server, lists its tools, calls them. |
| FastMCP ⚡ | Python framework atop the official mcp SDK — build servers in a few lines. |
| LangChain ↔ MCP 🤖 | The adapter that turns MCP tools into LangChain tools (client side). |
This repo is that stack in miniature: build servers (FastMCP) + use them (LangChain / raw SDK), then package & deploy them.
📚 Research — a tour of the folders
The learning ladder (suggested reading order):
🔬 deepdive-mcp learn the protocol + both kinds of client
🌦️ weather-node-mcp the same thing in TypeScript (not just Python)
📁 file-access-python-mcp build a real, clean FastMCP server
📦 mcp-server-deployment → package it as an installable library
🚪 mcp-server-gateway → aggregate many servers behind one endpoint
🐳 mcp-docker-deployment → ship it in a container
🤖 mcp-server-with-pydantic-ai-agent → consume it from the other side:
Pydantic AI agents, from raw JSON-RPC to a secured REST API
| 📁 Repo | What's inside | 🎯 Takeaway |
|---|---|---|
| deepdive-mcp 🔬 | Three parallel scenarios — ① raw mcp SDK over stdio, ② FastMCP over stdio + streamable-HTTP, ③ a LangChain MCP-adapter client hitting a community duckduckgo server. |
The 3 layers of MCP in one place: SDK server → FastMCP → LangChain client. The core “what / where / how” playground. |
| weather-node-mcp 🌦️ | Minimal MCP server in TypeScript using the official @modelcontextprotocol v2 client + server and zod. One tool (get-alerts, US NWS) + one resource, and a demo client. |
MCP in the other language — and that server and client can live in the same folder. |
| file-access-python-mcp 📁 | Production-flavoured FastMCP server: exposes a local folder of text docs as 9 tools (search / list / glob / get / edit / create / delete) over stdio and streamable-HTTP+CORS, with --allow-* permission flags. |
The cleanest example here — a solid template (domain / tools / transport split + e2e tests). Copy this one. |
| mcp-server-deployment 📦 | agent_terminal_tools: a FastMCP server of shell & file tools (run command, list, glob, grep, read/write/append/delete/copy/move) packaged as an installable Python package (uv build + entry point); client demos included, test suite still to come. |
How to turn a server into a package others can pip install / uvx. |
| mcp-server-gateway 🚪 | A FastMCP gateway: one server that mount(create_proxy(...))s several upstreams (DuckDuckGo + agent_terminal_tools) and re-exposes them over streamable-HTTP. |
The aggregate / proxy pattern — many servers behind a single endpoint. |
| mcp-docker-deployment 🐳 | The gateway above, containerised: python:3.14-slim + uv, pre-installs a uvx community server, runs it over streamable-HTTP. |
How to ship an MCP server as a Docker image. |
| mcp-server-with-pydantic-ai-agent 🤖 | An 8-step training ladder that builds to a finished app: 01 a hand-rolled JSON-RPC dispatcher (the wire format under MCP) → 02 the MCP protocol on it (initialize / tools/list / tools/call, JSON-Schema inputSchema) → 03 the same server in FastMCP + its client → then agents driving it: 04 a single tool that writes a haiku to a file (source of the loose *-haiku.txt in the folder root) → 05 a multi-tool loop (weather → web search → hotel selection) → 06 an orchestrator/worker pairing where researcher & actor agents are wrapped as tools of a manager → 07 the agent behind FastAPI (/health + /chat, uvicorn in a thread) → 08 the JWT & password-hash building blocks. main.py is the product: OAuth2 login, a JWT guard, the MCP calculate tool, and a self-test that exercises every auth failure. reference.py = the original Colab course the ladder was distilled from. |
The client-side companion to the server tour — how an agent consumes MCP tools, then productises into a secured REST API. |
🚧
weaviate-mcp/is a vendored Go MCP server for the Weaviate vector DB — WIP, not yet published, so it's left out of the tour.
📜 License & audit
- Licensed MIT — one repo-root license covers every sub-project.
- SECURITY_AUDIT.md records a security audit's findings
and this repo's verdict on each — including the
grep"handler mutates" claim, verified against the source and found to be a false positive, and the "0/33 tools in tests" claim, which overstates the real (partial) gap.
MCP Servers MCP questions
What is the Model Context Protocol (MCP) used for?
MCP is an open, JSON-RPC-based standard that allows LLM-powered applications to interact with the outside world through tools, resources, and prompts.
Can I build MCP servers in languages other than Python?
Yes, the repository includes examples of MCP servers built in TypeScript using the official modelcontextprotocol library.