# MCP Tool An **MCP tool** is a callable function exposed by an [MCP server](mcp-server) that an AI [agent](agent) can invoke to perform actions in the world. Tools are the primary mechanism through which agents extend their capabilities beyond text generation — enabling them to interact with external systems, APIs, file systems, and services. ## Anatomy of an MCP Tool Each MCP tool has: - **Name** — A unique identifier (e.g., `createTask`, `reply`) - **Description** — Natural language explanation of what the tool does (used by the LLM to decide when to call it) - **Input schema** — JSON Schema defining the parameters the tool accepts - **Response** — The data returned to the agent after the tool executes ## AgentRQ MCP Tools AgentRQ exposes the following tools to [Claude Code](claude-code) agents: | Tool | Parameters | Description | |------|-----------|-------------| | `createTask` | `title`, `description`, `status` | Create a task on the task board | | `reply` | `chat_id`, `message` | Send a message to the human | | `getTaskMessages` | `chat_id` | Retrieve conversation history | | `updateTaskStatus` | `task_id`, `status` | Set a task to ongoing or completed | | `downloadAttachment` | `attachment_id` | Retrieve a file attachment | | `getWorkspace` | — | Get workspace configuration | ## How the Agent Uses Tools When a tool is available, the LLM can choose to call it by generating a structured tool call in its response. The [MCP client](mcp-client) intercepts this, executes the tool call against the server, and returns the result to the model as a new context entry. ``` 1. Model generates: { "tool": "reply", "params": { "chat_id": "...", "message": "Done!" } } 2. MCP client calls AgentRQ's reply endpoint 3. AgentRQ delivers the message to the human's task board 4. MCP client returns: { "success": true } to the model 5. Model continues reasoning with updated context ``` ## Related Terms - [MCP](mcp) - [MCP Server](mcp-server) - [MCP Client](mcp-client) - [Tool Use](tool-use) - [Task](task)