mcp-use-ts
mcp-use is the framework for MCP with the best DX - Build AI agents, create MCP servers with UI widgets, and debug with built-in inspector. Includes client SDK, server SDK, React hooks, and powerful dev tools.
claude mcp add --transport stdio mcp-use-mcp-use-ts npx @mcp-use/server-mcp-use-ts \ --env GITHUB_TOKEN="your-token-if-needed (optional)"
How to use
MCP-Use TS is a TypeScript framework for building and consuming MCP servers and clients with an integrated inspector and UI widgets. The server side exposes MCP tools and resources that can be called by LLMs or agents, while the client side demonstrates how to connect to those servers, discover available tools, and invoke them. Use cases include building domain-specific MCP servers (for weather, finance, data pipelines, etc.) and creating UI widgets that render tool results within an inspector-enabled interface. The toolkit includes tooling to hot-reload during development and to bundle UI widgets for standalone deployment, making it straightforward to iterate on tool definitions and UI components. You can connect to multiple MCP servers from a single MCP client and control tool access and observability through the provided abstractions.
How to install
Prerequisites:
- Node.js (14+ or 16+ recommended) and npm/yarn installed
- Basic familiarity with TypeScript and npm package management
Installation steps:
-
Install Node.js from https://nodejs.org/ if you don’t already have it.
-
Install the MCP-Use TS server package (example uses npm):
npm init -y npm install --save @mcp-use/server-mcp-use-ts
-
Initialize or scaffold an MCP server project (if starting from scratch):
npx @mcp-use/cli create-mcp-use-app my-mcp-app cd my-mcp-app
-
Install dependencies for the server inside your project:
npm install
-
Run the development server with hot-reload and inspector:
npm run dev
-
Open http://localhost:3000 to access the MCP server UI and inspector. Adjust ports or configuration as needed in your project setup.
Additional notes
Tips and common considerations:
- The inspector UI is auto-mounted in development mode, typically at /inspector. Use this to debug tools and view tool calls.
- You can host MCP servers with multiple transports (HTTP/SSE or WebSocket) depending on your deployment needs.
- When defining tools, consider adding Zod schemas for parameters to enforce type safety and improve tooling support.
- Environment variables (e.g., GITHUB_TOKEN in examples) can be supplied via the mcp_config env section or your deployment platform’s secret management.
- If you publish or share your MCP server, ensure your authentication and OAuth flows (if used) are properly configured for secure access.
- The npm package name used in examples is @mcp-use/server-mcp-use-ts; adjust if your organization or package naming differs.
Related MCP Servers
mcp-router
A Unified MCP Server Management App (MCP Manager).
HyperChat
HyperChat is a Chat client that strives for openness, utilizing APIs from various LLMs to achieve the best Chat experience, as well as implementing productivity tools through the MCP protocol.
AgentChat
AgentChat 是一个基于 LLM 的智能体交流平台,内置默认 Agent 并支持用户自定义 Agent。通过多轮对话和任务协作,Agent 可以理解并协助完成复杂任务。项目集成 LangChain、Function Call、MCP 协议、RAG、Memory、Milvus 和 ElasticSearch 等技术,实现高效的知识检索与工具调用,使用 FastAPI 构建高性能后端服务。
mcp-graphql
Model Context Protocol server for GraphQL
python -client
支持查询主流agent框架技术文档的MCP server(支持stdio和sse两种传输协议), 支持 langchain、llama-index、autogen、agno、openai-agents-sdk、mcp-doc、camel-ai 和 crew-ai
cli
Fine-grained control over model context protocol (MCP) clients, servers, and tools. Context is God.