Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

Glossary

MCP agents

Agents that reach every tool and data source through MCP servers instead of custom code, so integrations stay separate from the agent's logic. The name also belongs to one framework, LastMile AI's open-source mcp-agent.

The MCP specification says hosts must obtain the user's explicit consent before invoking any tool, that there should always be a person able to deny a tool call, and that clients must treat tool annotations as untrusted unless the server is trusted. Every server an agent uses is one more dependency to approve, pin and watch.

Open the glossary in the Learning Center