{"slug": "stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript", "title": "Stop Using Fragile Chatbots: Build Standard MCP Agents with TypeScript", "summary": "A developer built an open-source TypeScript Agent Orchestrator that turns a natural-language prompt into a standard-compliant Agent Plugin, generating the plugin.json manifest, mcp.json tool configuration, and SKILL.md workflow files. The tool includes a platform interceptor that wraps npx/uvx commands in cmd /c so filesystem MCP servers run on Windows, plus a React/Vite web UI that parses intent, selects an MCP server, and estimates token cost before execution. It also ships a BudgetGuardian that halts the agent once API costs exceed $0.50 and requires explicit confirmation for destructive file operations.", "body_md": "In 2026, we are drowning in AI chatbots. Every SaaS has a \"Chat with our data\" button. But if you're a developer or an agency building for clients, you've probably hit a wall: LLMs are trapped in their chat boxes.\n\nTo make an LLM actually do things—like read a local file, trigger an API, or move a PDF—you either rely on brittle Zapier workflows or spend weeks writing custom OpenAI function-calling code.\n\nThat’s why MCP (Model Context Protocol) and the Agent Plugins 1.0 ecosystem were born. It’s the industry standard for giving models safe access to real tools.\n\nRecently, I built an Agent Orchestrator—an open-source, personal software factory. It takes a natural language prompt and generates a fully compliant Agent Plugin (MCP + Skills). Today, I’m pulling back the curtain on how it works and how you can use it to automate real business workflows.\n\nThe Architecture: Three Protocols, One Goal\n\nBuilding an AI agent isn't just about prompting. A standard-compliant agent directory needs three things:\n\nplugin.json: The Agent Plugins 1.0 manifest (metadata).\n\nmcp.json: The Model Context Protocol configuration (tools).\n\nSKILL.md: The Agent Skills instructions (workflows).\n\nHere is the exact directory structure my tool generates:\n\ntext\n\nagents/my-agent/\n\n├── plugin.json          # Manifest\n\n├── mcp.json             # Tools connection\n\n└── skills/my-agent/\n\n    └── SKILL.md         # Workflow instructions\n\nThe \"Aha!\" Moment: Solving the Windows MCP Bug\n\nIf you've tried running the official @modelcontextprotocol/server-filesystem on Windows, you know the pain.\n\nWindows doesn't recognize npx as an executable in stdio transport. It expects a .cmd shim. The standard fix is wrapping it in cmd /c.\n\nIn my TypeScript Orchestrator, I built an automatic platform interceptor:\n\ntypescript\n\nexport function normalizeMcpConfigForPlatform(config: any) {\n\n  if (process.platform !== \"win32\") return config;\n\n// Wrap npx/uvx for Windows compatibility\n\n  if (/^(npx|uvx|npm)$/i.test(config.command)) {\n\n    return {\n\n      command: \"cmd\",\n\n      args: [\"/c\", config.command, ...(config.args || [])]\n\n    };\n\n  }\n\n  return config;\n\n}\n\nNow, when the agent is created, the mcp.json is instantly ready for Claude Desktop or Cursor, even on Windows.\n\nEnter the Web UI: Drag, Drop, Generate\n\nCLI is great for power users, but to make this accessible, I built a sleek Web UI (React + Vite + Tailwind).\n\nYou type: \"Organize PDFs in my downloads folder by project name.\"\n\nThe UI hits the backend API, parses the intent, and dynamically selects the filesystem MCP server. It also calculates the estimated token cost (using DeepSeek or GPT-4o) before you even run it.\n\nWhy This Matters for Businesses\n\nIf you are a freelance developer, you know clients don't pay for \"AI.\" They pay for saved hours.\n\nWith a standardized MCP agent, you can now hand a client a folder. They double-click start.bat, and they have a local, secure AI assistant that manages their Excel sheets, categorizes their invoices, or drafts email replies—without sending sensitive data to a cloud database.\n\nWhat's Next?\n\nThe Agent Orchestrator is just the beginning. The true power of MCP lies in the ecosystem. We are moving from \"Chat\" to \"Action.\"\n\nCost Control: The tool includes a BudgetGuardian that kills the agent if API costs exceed $0.50.\n\nSecurity: Every destructive file operation requires explicit user confirmation.\n\nWant to build your own automated agent?\n\nI'm currently helping small businesses integrate MCP workflows to cut their operational costs.\n\n**Want to see how this works or need a custom automation solution for your business?**\n\nI help small businesses save hours of manual work by integrating local MCP agents. \n\n👉 **Let's discuss your project:** [Email me](mailto:haoyang20141949@outlook.com)", "url": "https://wpnews.pro/news/stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript", "canonical_source": "https://dev.to/mcpforge/stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript-1i8", "published_at": "2026-10-17 14:54:30+00:00", "updated_at": "2026-10-17 15:04:42.917696+00:00", "lang": "en", "topics": ["ai-agents", "agent-protocols", "ai-tools", "developer-tools"], "entities": ["Model Context Protocol", "Agent Plugins 1.0", "Claude Desktop", "Cursor", "DeepSeek", "GPT-4o", "React", "Vite"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript", "markdown": "https://wpnews.pro/news/stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript.md", "text": "https://wpnews.pro/news/stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript.txt", "jsonld": "https://wpnews.pro/news/stop-using-fragile-chatbots-build-standard-mcp-agents-with-typescript.jsonld"}}