you need to learn MCP RIGHT NOW!! (Model Context Protocol)

NetworkChuck Technology Careers Career & Money

About this video

Chuck explains the Model Context Protocol (MCP) as a standardised way to give an LLM tools without forcing it to write API code. He compares it to USB-C and notes it was created by Anthropic: an MCP server sits between the model and an application, hides the API calls, authentication and code, and exposes plain-language tools the model simply asks to use. Using Docker Desktop’s MCP Toolkit and its catalog, he connects an Obsidian server (which he says offers 12 tools) to Claude Desktop, so Claude creates and searches notes in his vault via Obsidian’s local REST API plugin. He then connects the same servers to LM Studio (trying local models including Gemma 3 and DeepSeek, with mixed results) and to Cursor, which fetches a YouTube transcript and files a summary in Obsidian. The second half is about building custom servers with a prompt he wrote: he has an LLM generate the files for a dice-roller server, registers it in a custom Docker catalog and registry, then repeats the process for a Toggl timer server (start, stop and view timers, with the API token kept in Docker MCP secrets) and for a Kali Linux hacking server, troubleshooting whitelist, secret and Dockerfile problems before running scans against a deliberately insecure test web app. Finally he explains the plumbing: local servers are Docker containers spun up only when called and talk over standard input/output using JSON-RPC messages, remote servers use HTTP/HTTPS and server-sent events, and Docker’s MCP Gateway lets one client connection reach many servers while centralising secrets. He runs the gateway over the network on port 8811 and connects it to n8n, where an agent chains restaurant search, Airbnb and Obsidian tools, and a remote CoinGecko server is added to Cursor to answer a crypto price question.

Key takeaways

MCP is a standard from Anthropic: an MCP server abstracts an app’s API, code and authentication, exposing plain-language tools an LLM can call without writing code.
Docker Desktop’s MCP Toolkit/catalog plus its gateway is presented as the easiest local route — add a server once, then connect clients such as Claude Desktop, LM Studio and Cursor.
Custom servers can be built by describing the job to an LLM with his build prompt: generate the Dockerfile, requirements and server files, build the container, then add it to a custom catalog and the registry.
Local MCP containers run only while a tool is being used and communicate over standard input/output with JSON-RPC; remote servers use HTTP/HTTPS with server-sent events.
The Docker MCP Gateway gives a client one connection to many servers, centralises secrets/API keys in Docker, and can be exposed over the network (he uses port 8811) for tools like n8n.

Episode details: published September 12, 2025 · 38:40 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are NetworkChuck’s own claims, shared for education — they are not typical or guaranteed.

← Previous: THIS Is Where the Internet Lives · Next: n8n Now Runs My ENTIRE Homelab →

All NetworkChuck videos · NetworkChuck profile · NetworkChuck on YouTube