How agents use tools, and where MCP fits
Separate the model, the agent loop, and the tool connection.
In plain language
An agent system can use a model to select a next step, call a tool, inspect the result, and continue toward a goal. The surrounding application sets boundaries such as available tools and stopping conditions.
MCP is a standard way for AI applications to connect to external capabilities and context. It is a connection protocol, not a language model or a guarantee that an action is safe. Tool execution still needs authentication, authorization and appropriate checks.
To help draft a news item, an agent might search sources, read evidence, draft a summary and request review. A separate publishing permission controls whether a draft can become public.
Try it yourself
Sketch a task with one read-only tool and one publishing tool. Mark the point where permission must be checked.
Go deeper · for developers and architects
Distinguish deterministic workflows from open-ended agent loops. Bound tool calls, cost and runtime. Log action outcomes and apply permissions at the tool boundary.
Check your understanding
2 related questions, with an explanation for each answer.
Report an issue
Tell the editors what needs correcting. Reports are reviewed and do not automatically remove content. Do not include personal or confidential information.
Finished this lesson?
Completion records your choice, not a measured skill level.