Build fluency in the vocabulary of connecting an AI assistant to tools through a standardized protocol.
0 / 5 completed
1 / 5
At standup, a dev mentions adopting a standardized protocol so any compliant AI assistant can connect to a company's internal tools and data sources without a custom integration for each one. What is this protocol called?
The Model Context Protocol, or MCP, is a standardized protocol that lets any compliant AI assistant connect to a company's internal tools and data sources through a common interface, rather than requiring a custom integration built separately for each different assistant. This dramatically reduces the integration work needed as the number of both tools and AI assistants grows. It plays a role similar to how a standardized API format reduces one-off integration work between unrelated systems generally.
2 / 5
During a design review, the team wants to expose a specific internal capability, like querying a database, as a discrete callable action an MCP-connected assistant can invoke. Which capability supports this?
Defining an MCP tool with a clear input and output schema exposes one specific internal capability, like a database query, as a discrete, individually callable action an MCP-connected assistant can invoke directly. Exposing the entire internal system with no discrete actions leaves the assistant with no clear, safe way to interact with it. This tool-definition pattern is what lets an assistant meaningfully act, not just read information, through the protocol.
3 / 5
In a code review, a dev notices the MCP server explicitly declares which resources and tools it exposes, rather than granting a connected assistant unrestricted access to everything on the host system. What does this represent?
Explicit capability declaration and scoping has an MCP server clearly define exactly which resources and tools it exposes, rather than granting a connected assistant unrestricted access to the entire host system by default. Unrestricted default access means a bug or malicious prompt could potentially cause far more damage than the intended integration required. This explicit scoping keeps an MCP integration's blast radius limited to only what was deliberately exposed.
4 / 5
An incident report shows an MCP-connected assistant was granted a write-capable tool for an integration that only ever needed to read data, and a bug caused an unintended write. What practice would prevent this?
Exposing only the minimum tool capability actually needed, like read-only access for an integration that never needs to write data, limits what a bug or unexpected behavior can actually cause. Exposing every available capability by default regardless of actual need unnecessarily widens that risk. This minimum-necessary-capability principle applied to MCP tool exposure is a direct application of least privilege to this specific integration pattern.
5 / 5
During a PR review, a teammate asks why the team adopts the Model Context Protocol instead of building a custom integration directly between their internal tools and each specific AI assistant. What is the reasoning?
A custom integration built separately for each different AI assistant means that same integration work gets repeated and maintained independently as more assistants and tools are added over time. A standardized protocol lets any compliant assistant connect through one common interface, avoiding that repeated effort. The tradeoff is depending on the protocol's own evolving standard and the maturity of client and server implementations that support it.