vision-one-mcp-server CLAUDE.md

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

Trend Vision One MCP Server - A Go-based Model Context Protocol (MCP) server that bridges AI tooling (Claude, VSCode + GitHub Copilot) with Trend Vision One security platform APIs. Enables natural language interaction with security services like Workbench alerts, Cloud Posture, endpoint management, attack surface discovery, and AI security guardrails.

Build & Test Commands

# Build
make mcpserver              # Build to ./bin/v1-mcp-server
go build -o ./bin/v1-mcp-server ./cmd/v1-mcp-server/main.go

# Test
go test -v ./...            # Run all tests

# Lint & Format
./script/check-gofmt        # Check formatting
./script/lint               # Run golangci-lint
gofmt -s -w ./              # Auto-format code

# Run locally
./bin/v1-mcp-server -region us  # Requires TREND_VISION_ONE_API_KEY env var

Architecture

cmd/v1-mcp-server/main.go     # Entry point, CLI flags, region validation
internal/v1mcp/server.go      # MCP server setup, tool registration
internal/v1mcp/tools/*.go     # Tool handlers (one file per domain)
internal/v1client/*.go        # HTTP client, API endpoint methods

Request Flow: MCP Request → Tool Handler → Parameter Extraction → v1client API call → HTTP → Response → MCP Result

Key Conventions

Tool Implementation Pattern

Each tool is a factory function returning mcpserver.ServerTool:

func toolDomainResourceAction(client *v1client.V1ApiClient) mcpserver.ServerTool {
    return mcpserver.ServerTool{
        Tool: mcp.NewTool("domain_resource_action",
            mcp.WithDescription("..."),
            mcp.WithString("param", mcp.Description("...")),
            mcp.WithToolAnnotation(mcp.ToolAnnotation{ReadOnlyHint: toPtr(true)}),
        ),
        Handler: func(ctx context.Context, request mcp.CallToolRequest) (*mcp.CallToolResult, error) {
            // Extract parameters, call API, return result
        },
    }
}