Model Context Protocol for AI Engineers
Model Context Protocol (MCP) is an open protocol that lets AI applications connect to external tools, data and prompts through a standard client-server interface. An MCP server exposes capabilities, and a host application reaches them through a client. On the AI Engineer path it replaces one-off integrations with reusable servers that must be secured carefully.
Stage: Agents and Tool Use · Part of the AI Engineer roadmap · Last reviewed 19 November 2026
What to learn in Model Context Protocol
- Explain the roles of host, client and server in an MCP connection
- Build an MCP server that exposes one tool and one resource
- Choose between stdio and HTTP-based transports for a given deployment
- Apply least privilege and user confirmation to risky MCP tools
- Treat tool descriptions and results from third-party servers as untrusted input
Model Context Protocol interview questions
What is the difference between an MCP tool, a resource and a prompt?
A tool is a function the model can invoke to take an action or compute something. A resource is read-only data the application can load into context, identified by a URI. A prompt is a reusable template a user can pick. Tools are model-controlled, resources application-controlled and prompts user-controlled.
What is the tool poisoning risk with third-party MCP servers?
A malicious or compromised server can hide instructions in a tool description or result, steering the model to leak data or call other tools. Mitigate by vetting servers, pinning versions, showing users each tool call for approval and limiting every server's permissions.
Where Model Context Protocol fits in the AI Engineer roadmap
Model Context Protocol belongs to the Agents and Tool Use stage (intermediate level), which takes about 84 hours. Software that acts.
Other units in this stage
Back to the full AI Engineer roadmap