What are MCPs?
MCPs (Model Context Protocol) are essentially a way to expose tools and context to agents in a standardized way, so they can extend their capabilities and interact with different systems without having to implement custom integrations for each one of them.
TIP
Anyone can implement a MCP server, although service providers (e.g. GitHub) often have an official implementation.
Where is useful
MCPs really shine when you need multiple tools or contexts involved in a given process or workflow. For example, if you have one task that needs to talk with Service A, another Service B, and then do some things on Service C, you could streamline that process by delegating this task to an agent that have these services exposed as tools on a MCP server.
Most developers, including myself, struggle to adopt MCP on a day to day basis, because we default to the tools we know like CLIs and so on, but the reality is that if your work needs to involve multiple systems as mentioned before, MCPs are really usefull.
For example, you could have an issue created on GitHub and a task in Jira while a message is sent to Slack in the same prompt.