Definition
A tool call is a structured request that invokes a named function with arguments and returns a result or an error to the calling application.
What happens during a tool call
An MCP client first reads the tools exposed by a server. Each tool has a name, description and input schema. When the model decides a tool is relevant, the client sends the selected tool name and structured arguments to the server. The server validates the request, performs the work and returns content, structured data or an error.
This structure lets the application distinguish a market lookup from an order request and lets the server reject missing or invalid arguments before running the underlying action.
A request is not a completed action
A tool call can read data, prepare an operation or request execution. It may fail validation, time out or be rejected by an account policy or trading venue. The result, not the call alone, shows whether the requested operation completed.
What it does not mean
A tool call is not automatically a trade. MCP tools can retrieve information, change settings or submit actions, depending on what the server exposes.
Where this term fits
Questions about tool call
- Who chooses when an MCP tool is called?
- Tools are model controlled in the MCP control model, while the client mediates the request and the server decides whether and how to execute it.
- What does a tool call contain?
- It identifies a tool and supplies arguments that should match the tool's declared input schema. The returned result can contain text or structured content.
- Can a tool call fail?
- Yes. Invalid arguments, account restrictions, unavailable services and venue errors can all prevent the requested operation from completing.
See these concepts in context
Review the Fensory MCP connection, tools and account controls.