What is MCP?
Purpose: explain the Model Context Protocol and why it changes the trust model for coding assistants
Official/source link: https://modelcontextprotocol.io/docs/getting-started/intro
Maintainer: Model Context Protocol project
License: Requires verification at implementation level
Last verified: 2026-07-14
Status: Documentation verified
Verification scope: official MCP docs and specification reviewed; no live server connected from this repository
Purpose
MCP is an open protocol for connecting AI applications to external systems such as files, APIs, databases, browsers, or custom tools.
Why It Matters
An ordinary prompt can only use the context you paste into the model. An MCP-enabled agent can ask external systems for more context or trigger actions through exposed tools. That makes the protocol powerful and risky at the same time.
Typical Use Cases
- repository and issue access
- browser automation
- database queries
- ticketing workflows
- internal search and retrieval
Core Risk
Once a host connects to an MCP server, the agent may be able to:
- read private data
- write to external systems
- execute privileged operations indirectly
- exfiltrate data through tool calls or logs
Evaluation Questions
- who maintains the server
- what credentials it requires
- what tools it exposes
- whether it can write, delete, or execute
- where data travels
- how access is revoked
Least-Privilege Starting Point
Start with:
- a read-only server if possible
- the narrowest credentials available
- a sandbox or test workspace
- explicit human approval for destructive actions
Verification Status
MCP itself is well documented, but individual server implementations vary widely in quality and risk. This repository only treats category pages as evaluation guides unless a specific implementation is responsibly verified.