Package ai.tabforge.workshop

Class StartReviewTool

java.lang.Object
ai.tabforge.workshop.StartReviewTool
public class StartReviewTool extends Object
MCP Tool: start_review

Entry point for the entire review workflow. Claude calls this tool when the developer asks to review their code, e.g. "Review my changes before PR" or "Scan the whole project". The tool creates a ReviewSession, launches all sub-agents in a background thread, and immediately returns a reviewId — without waiting for the review to complete.

Flow: 


 Developer: "Review my changes"
     ↓
 Claude AI → calls start_review(project_path, scope)
     ↓
 StartReviewTool.execute() → creates ReviewScope
                           → calls OrchestratorAgent.startReview()
                           → returns { reviewId: "abc-123", status: "RUNNING" }
     ↓
 Claude AI → calls GetReportTool(reviewId) periodically to track progress

Created by: WorkshopServer.start() at server startup. Receives OrchestratorAgent via constructor injection.

CERTIFICATION NOTE — covers two exam domains:

  • Domain 4 — Tool Design & MCP Integration (18%): This class IS an MCP tool. The DESCRIPTION field is the instruction to Claude AI about when and how to call this tool. The INPUT_SCHEMA defines the JSON contract Claude must follow. Together they demonstrate Task Statement 4.1: "Design MCP tool interfaces with clear descriptions and input schemas."
  • Domain 1 — Agentic Architecture & Orchestration (27%): This tool is the trigger that starts the agentic loop. It demonstrates the async fire-and-forget pattern: return immediately with a handle (reviewId), let the orchestrator run agents in the background. This is Task Statement 1.7: "Manage session state, resumption, and forking."

  • Constructor Details

  • Method Details

    • toolSpecification

      public io.modelcontextprotocol.server.McpServerFeatures.AsyncToolSpecification toolSpecification(io.modelcontextprotocol.json.McpJsonMapper jsonMapper)
      Builds the MCP tool descriptor and attaches the async execution handler.

      The tool descriptor contains the tool name, natural-language description (shown to the AI model), and the JSON Schema for the input parameters. The execution handler receives the parsed arguments and delegates to

      invalid reference 
      #execute(String)
      .

      Called from: start() method of the main MCP server class once, during server initialization.

      Analogy: toolSpecification() is like registering a service at the hotel reception — you say: "We offer room service, you call us at the number 101, and you need to tell us the room number and what you want." The MCP server remembers that and waits for calls.
      Parameters:
      jsonMapper - MCP JSON mapper used to parse the INPUT_SCHEMA string into a schema object understood by the MCP framework Its like a translator between the Java and JSON worlds within the MCP framework. So: INPUT_SCHEMA (String) → jsonMapper → McpSchema object understood by the MCP framework.
      jsonMapper -
      Returns:
      a fully configured async tool specification ready to be registered with the MCP server