MCP Tools
One of the three primitives exposed by an MCP Server, representing executable actions such as creating an issue or running a database query. The model decides autonomously when to invoke them, making this the core mechanism by which MCP enables AI agents to perform external operations.
Timeline (last 90 days)
Tool descriptions are read verbatim by the model and are part of the prompt rather than ordinary documentation; the tool poisoning attack precisely hides malicious instructions in tool descriptions.
The Tool workflow is model-driven: the user's question is first sent to the LLM, which determines which tool to call, with decision-making authority residing in the model
The Resource mode is faster than the Tool mode because it eliminates the round-trip step of having the LLM first make a tool selection decision
The core of a tool is execute, while the core of a resource is read
All Facts (4)
Tool descriptions are read verbatim by the model and are part of the prompt rather than ordinary documentation; the tool poisoning attack precisely hides malicious instructions in tool descriptions.
50%UnverifiedThe Tool workflow is model-driven: the user's question is first sent to the LLM, which determines which tool to call, with decision-making authority residing in the model
50%UnverifiedThe Resource mode is faster than the Tool mode because it eliminates the round-trip step of having the LLM first make a tool selection decision
50%UnverifiedThe core of a tool is execute, while the core of a resource is read
50%