ModelContextProtocol
title: Model Context Protocol (MCP) type: platform created: 2026-05-16 last_updated: 2026-05-16 related: ["Playradar"] sources: ["https://www.devshorts.in/p/how-to-build-your-own-mcp-server"] radar_quadrant: Platforms radar_ring: Assess radar_position: inner
An open-source protocol developed by Anthropic that standardises how external applications provide context to large language models. MCP defines a client-server architecture enabling LLMs to connect to tools, data sources, and services through a uniform interface, regardless of which LLM provider is used.
MCP follows a client-server model with four components:
| Component | Role |
|---|---|
| MCP Host | The LLM application (e.g., Claude Desktop) that initiates connections |
| MCP Client | The connector managing communication between host and servers |
| MCP Server | Code exposing specific capabilities via the protocol |
| Data Sources | Local files, databases, or remote APIs the server wraps |
MCP servers expose three primitive types:
- Resources — Read-only data endpoints, analogous to GET requests. Return structured data to the LLM on request.
- Tools — Computational functions that perform actions or modifications, analogous to POST requests. Examples: send email, create calendar event, write file.
- Prompts — Reusable templates that guide LLM interactions for specific workflows.
Pre-built MCP servers are added to a host application by editing a configuration file (claude_desktop_config.json for Claude Desktop) with the server's path and parameters. The host discovers and connects to servers at startup.
Custom servers are built using the official Python or TypeScript SDK. A minimal server defines resource, tool, and prompt handlers using decorator syntax, then registers itself in the host configuration.
At time of writing, a growing catalogue of pre-built MCP servers covers: filesystem access, GitHub, Gmail, Google Calendar, databases, web search, and more. Claude Desktop provides native MCP support with full feature coverage across resources, tools, and prompts. Other clients support MCP with varying feature subsets.
MCP addresses three recurring problems in LLM integration work:
- Standardisation — Eliminates bespoke integration code per tool per model.
- Provider portability — The same MCP server works with any compliant LLM host, avoiding lock-in.
- Data locality — Servers run within the user's infrastructure; data does not need to leave the local environment to reach the LLM.
Model Context Protocol (MCP) sits in the Assess ring of the Platforms quadrant, at inner position. First studied via the devshorts.in article on 2026-05-16; no personal implementation to date. MCP is an Anthropic-originated open standard with rapidly growing ecosystem adoption — pre-built servers cover the most common integration targets and Claude Desktop supports it natively. Inner position reflects low friction to trial (adding a pre-built server requires editing one JSON config file) and strong applicability to any workflow involving LLM access to local files, APIs, or services. The remaining gate before Trial is a completed integration in an active project.
