MyMCP
A local, no-code builder for MCP servers that lets you create AI tools by describing them in plain language, without writing code. Connect your tools to any MCP-compatible assistant like Claude, Cursor, or VS Code Copilot.
README
<div align="center">
MyMCP
Build your own MCP servers — no code required.
A local, no-code builder for Model Context Protocol (MCP) servers. Assemble your own AI tools and connect them to Claude, Cursor, VS Code Copilot, and any other MCP-compatible assistant — without writing a single line of code.
Quick Start · Features · How it works · Connect your AI · Architecture · Contributing
</div>
What is MyMCP?
The Model Context Protocol (MCP) is the open standard that lets AI assistants such as Claude securely call external tools and APIs. Building an MCP server normally means writing TypeScript or Python, defining JSON schemas, and wiring up a runtime.
MyMCP removes the code. It is a local web app where you:
- Describe a tool in plain language.
- Add the input fields your tool needs.
- Point it at an API, an integration, or a script.
- Click Connect — and your AI assistant can use it.
It is built for people who know what they want their AI to do, but would rather not become backend engineers to make it happen: manual testers, product and project managers, business analysts, team leads, and AI power users.
Everything runs locally. No accounts, no cloud, no SaaS — your tools and secrets never leave your machine.
Features
- No-code Tool Wizard. Create AI tools step by step: name, AI-facing description, input fields, action. The JSON Schema is generated for you.
- Three ways to power a tool. Call any REST API, reuse a saved integration, or run a JavaScript / Python script for advanced cases.
- Built-in integrations. Jira, GitHub, Slack, Notion, and generic REST, with stored credentials and connection testing.
- One-click connect. MyMCP detects installed clients (Claude Desktop, Cursor, VS Code / Copilot) and writes the config for you. Existing servers are preserved and the previous config is backed up first.
- Manual export. Copy ready-to-paste config snippets for any MCP client.
- Git-friendly projects. Your entire setup is a single
project.jsonfile you can version, share, and import. - Local and private by design. No sign-up, no telemetry, no cloud. Secrets stay on your machine.
- One command to launch.
start.bat/start.shinstalls everything, boots the server and UI, and opens your browser.
Presentation
https://github.com/user-attachments/assets/aa10356a-446b-44bc-9ba0-1c5ab56ba132
Quick Start
Prerequisite: Node.js 18+ (tested on Node 22).
1. Clone the repository
git clone https://github.com/damiankaniewski/MyMCP.git
cd MyMCP
2. Run the launcher
Windows
start.bat
macOS / Linux
./start.sh
On first run the launcher will:
- Install all dependencies (root, backend, frontend).
- Start the backend on
http://localhost:3001. - Start the web UI on
http://localhost:5173. - Open your browser automatically.
How it works
MyMCP provides a small set of focused screens:
| Screen | Purpose |
|---|---|
| Dashboard | See your project, tools, and MCP status. Import / export project.json. |
| New tool | A step-by-step wizard: name, AI description, input fields, action. |
| Integrations | Connect Jira, GitHub, Slack, Notion, or any REST API (with secrets). |
| Connect to AI | Copy ready-made config, or auto-connect to Claude Desktop, Cursor, VS Code. |
Tool actions
When you create a tool, you choose what it does:
- Call an API — make an HTTP request. Insert field values with
{{fieldName}}and secrets with{{secrets.KEY}}. - Use an integration — reuse a saved connection; auth headers are added for you.
- Run a script — JavaScript or Python (advanced). Your code receives a
paramsobject and returns the result.
Example tool
A "Get Public Holidays" tool, built entirely in the UI, compiles down to this in your project.json:
{
"name": "get_public_holidays",
"description": "Returns public holidays for a given country and year.",
"inputSchema": [
{ "name": "country", "type": "text", "required": true, "description": "Two-letter code, e.g. PL" },
{ "name": "year", "type": "number", "required": true, "description": "Year, e.g. 2026" }
],
"executionType": "http",
"executionConfig": {
"method": "GET",
"url": "https://date.nager.at/api/v3/PublicHolidays/{{year}}/{{country}}"
}
}
Your AI assistant now has a get_public_holidays tool it can call on its own.
Connect your AI assistant
Open the Connect to AI screen. There are two ways to connect.
Automatic (recommended)
MyMCP detects which apps are installed and shows a Connect button. One click writes the server straight into that app's MCP config — your existing servers are kept and the previous file is backed up first. Then restart the app (for VS Code, open Copilot Chat in Agent mode).
Manual
Pick the app's tab, click Copy, and paste the snippet into its MCP config file:
{
"mcpServers": {
"my-mcp": {
"command": "node",
"args": ["generated/server.mjs"]
}
}
}
Supported clients: Claude Desktop, Cursor, VS Code / Copilot, and any client that speaks MCP over stdio.
Architecture
mymcp/
├── frontend/ React + TypeScript + Vite + TailwindCSS web UI
├── backend/ Node + TypeScript + Express + official MCP SDK
│ └── src/
│ ├── generator/ turns project.json into a standalone MCP server
│ ├── routes/ project · tools · integrations · server · export
│ └── storage/ project persistence
├── projects/ your project.json lives here (Git-friendly)
├── templates/ starter integration templates (Jira, GitHub, Slack)
├── generated/ the compiled MCP server artifact (generated/server.mjs)
├── start.bat one-click launcher (Windows)
└── start.sh one-click launcher (macOS / Linux)
Tech stack
| Layer | Technologies |
|---|---|
| Frontend | React 18 · TypeScript · Vite · TailwindCSS · React Router · lucide-react |
| Backend | Node.js · TypeScript · Express · @modelcontextprotocol/sdk · Zod |
| Generator | Compiles project.json into a self-contained server.mjs (stdio) |
Data storage
Everything lives in a single, Git-friendly file:
projects/example-project/project.json
Export it from the Dashboard to back it up or share it. Local secrets are kept in a separate secrets.json that is git-ignored by default.
Developer commands
npm run install:all # install everything (root + backend + frontend)
npm run dev # run backend + frontend together with live reload
npm run build # type-check and build both
- Backend —
backend/(Node + TypeScript + Express + official MCP SDK). - Frontend —
frontend/(React + TypeScript + Vite + TailwindCSS). - Generator — turns
project.jsoninto a self-contained MCP server atgenerated/server.mjs.
The generated server links to the installed dependencies, so a connected client can launch it directly. To run it on another machine, run npm install inside generated/ first.
Roadmap
Shipped
- [x] No-code Tool Wizard (HTTP, integration, JavaScript / Python script)
- [x] One-click connect for Claude Desktop, Cursor, VS Code (Copilot)
- [x] Project import / export (
project.json) - [x] Connection presets with built-in auth (Jira, GitHub, Slack, Notion, REST)
- [x] Single-command launcher (backend + web app in one window)
Next — three pillars
1. Integration presets with ready-made tools Turn presets from connection-only into one-click packs that drop in working tools, so a new user gets value without building anything by hand.
- [ ] Bundle starter tools per service (e.g. Jira: search / create / comment on issues; GitHub: list / create issues, search repos)
- [ ] Add Google (Sheets, Calendar, Gmail) and broaden existing packs
- [ ] "Add pack" UX: pick a service, get its tools pre-wired to your credentials
- [ ] Friendly per-field guidance (where to get the token, scopes needed)
2. HTTP transport with auth (self-hosting) Run the generated server on a shared/corporate host instead of only locally over stdio.
- [ ] Streamable-HTTP / SSE transport option in the generator (alongside stdio)
- [ ] Auth on the HTTP endpoint (bearer / API key), configurable per server
- [ ] Remote-URL connect snippets for Claude, Cursor, VS Code
- [ ] Deploy notes (port, reverse proxy, secrets) for a corp server
3. Python scripting out of the box Make Python tools reliable on any machine — no "install Python first".
- [ ] Managed / detected Python runtime so scripts run with zero setup
- [ ] Per-tool dependencies (pip) resolved automatically
- [ ] Hardened execution wrapper (clean params in, structured result out)
Not planned: a public marketplace of community templates. The focus is curated, first-party presets that work out of the box.
Have an idea? Open an issue.
Contributing
Contributions, bug reports, and feature requests are welcome.
- Fork the repository and create your branch:
git checkout -b feature/my-feature - Make your changes and run
npm run buildto type-check. - Commit, push, and open a pull request.
The codebase favors modular, well-typed, feature-based code and a plugin-style architecture, so new integrations require minimal changes elsewhere.
Contributors
- Damian Kaniewski — @damiankaniewski
License
Distributed under the MIT License. See LICENSE for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。