MCP Tool

The name field is a unique, machine-readable string identifier for the tool within the server's namespace. It follows a reverse domain-like naming convention (e.g., github.search_issues) to ensure global uniqueness and avoid collisions when multiple servers are connected. This name is used by the client in the JSON-RPC tools/call request to specify which function to execute.

The description is a human-readable, natural language explanation of the tool's purpose and behavior. It is critically consumed by the AI client's language model to understand when and how to use the tool. A well-written description includes:

• The tool's primary action (e.g., 'Searches for repositories').
• The expected outcome.
• Any important preconditions or side effects. This text directly influences the agent's reasoning and tool selection accuracy.

The inputSchema is a JSON Schema object that rigorously defines the structure, types, and validation rules for the tool's arguments. This schema acts as a contract between the client and server, enabling:

• Structured Output Guarantees: The LLM is constrained to generate arguments that match the defined properties (e.g., query of type string).
• Request/Response Validation: The server validates incoming requests against this schema before execution.
• Automatic Documentation: Clients can introspect the schema to understand required and optional parameters, their data types, and any constraints (like enums or pattern matching).

The handler is the server-side function or method that contains the business logic executed when the tool is invoked. It is not part of the definition sent to the client but is its essential counterpart on the server. The handler:

• Receives the validated input arguments.
• Executes the intended action (e.g., calls a REST API, queries a database, runs a script).
• Returns a result (or error) that conforms to the MCP result specification.
• Manages authentication, error handling, and side effects securely within the server's execution environment.

During the initial capability negotiation handshake, the MCP server advertises its available tools to the client via the tools list in the initialize result. Each entry in this list is the complete tool definition (name, description, inputSchema). This allows the client to dynamically discover and register the tool's specification, enabling the AI to plan its use without prior hard-coded integration. The set of tools can be static or change dynamically based on server state or user permissions.

The formal specification of an executable function within the Model Context Protocol. A Tool Definition is a JSON object containing:

• name: A unique identifier for the tool.
• description: A natural language explanation of the tool's purpose.
• inputSchema: A JSON Schema object that strictly defines the structure, data types, and validation rules for the tool's input parameters. This definition acts as a contract between the server and client, enabling automatic discovery and safe invocation.

The server-side implementation function that contains the core business logic executed when a tool is invoked. Upon receiving a valid Tool Invocation request, the MCP server calls the corresponding Tool Handler. This function:

• Processes the validated input arguments.
• Executes the intended action (e.g., calling an external API, querying a database, running a script).
• Returns a structured result or error. The handler is where the protocol's declarative interface meets concrete execution code.

The act of an MCP client requesting the execution of a defined tool. A Tool Invocation is a JSON-RPC request sent from the client to the server containing:

• The name of the tool to call.
• An arguments object conforming to the tool's inputSchema. The server performs Schema Validation on the arguments, executes the associated Tool Handler, and returns a response with the result or an error. This is the core operational request in the MCP workflow.

The component within an MCP server responsible for managing tools. A Tool Provider is an interface that:

• Advertises the available Tool Definitions to the client during Capability Negotiation.
• Routes incoming Tool Invocation requests to the correct Tool Handler.
• Manages the execution context and lifecycle for tool calls. It serves as the central registry and dispatcher for all executable functions exposed by the server.

The process of verifying that data conforms to a defined structure before processing. In MCP, Schema Validation is critical for safety and reliability:

• Before a Tool Handler is executed, the server validates the invocation's arguments against the tool's inputSchema.
• This ensures type safety (e.g., numbers are numbers, required fields are present), prevents malformed requests from reaching business logic, and is a primary defense against prompt injection attacks that attempt to manipulate tool inputs.

The lightweight remote procedure call protocol that underpins all MCP communication. JSON-RPC (JavaScript Object Notation Remote Procedure Call) defines the message format for requests, responses, and notifications between MCP clients and servers. Key aspects include:

• Structured Messages: Every interaction is a JSON object with fields like jsonrpc, method, id, and params.
• Request-Response Model: Clients send requests with a unique id; servers send back a corresponding response.
• Transport Agnostic: It can be carried over Stdio Transport, SSE Transport, or other layers, with Message Framing ensuring reliable parsing.