API MCP Tools List and Call Results

Status: public · Confidence: medium (0.825) · Basis: verified_sources

## TL;DR

MCP tool discovery and call results are core API evidence for agents because they define what the model can invoke and what came back.

## Core Explanation

An agent cannot safely call tools from memory. It needs the current tools/list response, including names, descriptions, input schemas, annotations, and any output schema. After invocation, it needs the tools/call request, arguments, returned content, structuredContent, isError status, and any linked resources.

This is also a security boundary. A useful audit trail should show what tool list was exposed to the model, which arguments were shown to the user or host policy, and how the result was validated before it influenced follow-up reasoning.

## Source-Mapped Facts

- The Model Context Protocol tools specification says servers can expose tools that language models can invoke. ([source](https://modelcontextprotocol.io/specification/2025-06-18/server/tools))
- The Model Context Protocol tools specification says clients send a tools/list request to discover available tools. ([source](https://modelcontextprotocol.io/specification/2025-06-18/server/tools))
- The Model Context Protocol schema reference defines CallToolResult as the server response to a tool call, including content and optional structured content. ([source](https://modelcontextprotocol.io/specification/2025-06-18/schema))

## Further Reading

- [MCP Tools](https://modelcontextprotocol.io/specification/2025-06-18/server/tools)
- [MCP Schema Reference](https://modelcontextprotocol.io/specification/2025-06-18/schema)