io.github.prompeteer/prompeteer-mcp
编码与调试by prompeteer
Generate expert AI prompts for 140+ platforms with 16-dimension quality scoring.
什么是 io.github.prompeteer/prompeteer-mcp?
Generate expert AI prompts for 140+ platforms with 16-dimension quality scoring.
README
Prompeteer MCP Server
Contextual prompts and agent skills for 140+ AI platforms.
Prompeteer generates contextual prompts and agent skills tuned to the model you are actually using, across 140+ AI platforms and 129 languages. Every prompt gets a 16-dimension Prompt Score, and skills get a Skill Score. Saved prompts live in PromptDrive and roam across platforms. Free to start.
Connect to the hosted server
| Setting | Value |
|---|---|
| Endpoint | https://prompeteer.ai/mcp |
| Transport | Streamable HTTP |
| Authentication | OAuth 2.1 with PKCE S256 |
| Connection guide | prompeteer.ai/connect |
Use a client that supports remote MCP and OAuth discovery. Add the endpoint in its remote-server settings and complete the browser sign-in. Access depends on the account, plan, consented scopes, and client support.
For Cursor, add this entry to .cursor/mcp.json:
{
"mcpServers": {
"prompeteer": {
"url": "https://prompeteer.ai/mcp"
}
}
}
Other clients have their own configuration formats. Use the connection guide for client-specific setup.
Connect through the npm stdio bridge
For clients that launch local MCP commands, use Node.js 22.14.0 or later and add:
{
"mcpServers": {
"prompeteer": {
"command": "npx",
"args": ["-y", "@prompeteer.ai/mcp-server@2.0.3"]
}
}
}
The command runs a stdio bridge to the same hosted Prompeteer endpoint. It uses the pinned mcp-remote 0.8.6 proxy for Streamable HTTP and OAuth. The client opens a browser for sign-in when authorization is needed. The local bridge does not run Prompeteer's generation models or product service.
The endpoint and OAuth resource are fixed to https://prompeteer.ai/mcp. The wrapper uses a loopback OAuth callback, runs without a shell, and forwards MCP messages through stdout. Connection diagnostics use stderr. It does not accept arbitrary endpoints, authentication headers, or debug flags. Native remote clients can use the direct URL without installing the bridge.
npx @prompeteer.ai/mcp-server@2.0.3
npx @prompeteer.ai/mcp-server@2.0.3 --help
npx @prompeteer.ai/mcp-server@2.0.3 --json
npx @prompeteer.ai/mcp-server@2.0.3 --version
--json prints the direct remote configuration for Cursor. --help and --version do not start a connection.
OAuth credentials are managed by mcp-remote, using its local ~/.mcp-auth cache by default. On POSIX systems, new credential files are written with mode 0600. The bridge does not add its own token store. The package includes the tested proxy runtime and the patched query parser used by its callback server. Client installation does not resolve a different proxy dependency tree.
Capabilities
| Tool family | Examples |
|---|---|
| Prompt and skill generation | generate_prompt, generate_skill |
| Prompt Score | score_prompt |
| PromptDrive | list_prompts, get_prompt, save_to_promptdrive |
| Private Memory | list_memory_sources, search_memory, get_memory_page, ask_memory, get_memory_settings, update_memory_settings |
Skills receive a Skill Score as part of the skill workflow; score_skill is not a separate advertised MCP tool. Discover the current tool schemas through the authenticated server. Generation, scoring, writes, and private Memory operations may require additional scopes or consume plan allowances.
Example: “Generate a prompt for a product launch email in Claude. Show its Prompt Score, then save the approved version to PromptDrive.”
Authentication
The server advertises OAuth metadata and requests the scopes needed for authorized work. Keep credentials in the MCP client's managed authentication flow.
| Scope | Purpose |
|---|---|
mcp:read | Read prompts and vault metadata |
mcp:write | Save prompts |
mcp:generate | Generate prompts and skills |
memory:read | Read private Memory |
memory:write | Update Memory settings |
OAuth metadata and protected resource metadata are publicly discoverable. An unauthenticated MCP request returns HTTP 401 until you sign in.
Pricing and public facts
Free to start. See current plans and allowances.
Prompeteer reports 12,000+ users across 160 countries and supports 129 languages. These company facts were approved for publication in September 2026.
Registry and source
The official MCP registry lists the server as io.github.prompeteer/prompeteer-mcp. This is separate from first-party connector directories.
server.json describes both the npm stdio bridge and the hosted Streamable HTTP endpoint. Its registry description uses the canonical tagline because the registry limits descriptions to 100 characters. The full description is preserved in publisher metadata. .mcp/server.json is an identical compatibility copy.
Version 2.0.3 adds a working stdio bridge and corrects stale setup instructions. Earlier npm 2.0.0 printed setup instructions and exited, even though registry versions 2.0.0 through 2.0.2 advertised it as a stdio transport. The hosted endpoint remains https://prompeteer.ai/mcp.
Repository manifests for Cursor and Gemini CLI
plugin.json and mcp.json provide the portable Agent Plugins format supported by Cursor. After installing the plugin through Cursor, connect Prompeteer and complete browser sign-in. The direct Cursor configuration above also works as a manual setup format. Generated skills are returned as content; connecting does not run or install a generated skill.
gemini-extension.json provides a Gemini CLI extension using the hosted HTTP endpoint and OAuth discovery. Once the manifest is available in the public repository, install with:
gemini extensions install https://github.com/prompeteer/prompeteer-mcp
Restart Gemini CLI and complete the Prompeteer connection when prompted. Generation and scoring can consume account quota; requested saves write to your PromptDrive.
These manifests have been structurally validated. Installation and OAuth in the actual Cursor and Gemini CLI applications have not been tested for this release. Their presence does not establish a marketplace listing or platform approval. See the Cursor plugin documentation and Gemini extension reference.
Plugin source for Claude Code, Cowork and Grok Build
.claude-plugin/plugin.json supplies plugin metadata and .mcp.json connects the client directly to https://prompeteer.ai/mcp over HTTP. This plugin contains one remote MCP connection. It adds no local command, hook, skill, agent, or shell execution. It does not start the npm bridge or install its runtime dependencies.
For a local Claude Code check after cloning this repository:
claude plugin validate .
claude --plugin-dir .
Use /mcp in Claude Code to complete the Prompeteer OAuth connection. For Grok Build, the repository can be installed through a marketplace entry that pins its exact public commit. Marketplace submission, listing, and functional client testing are separate outcomes; these source files alone do not establish any of them.
Network access and account permissions
The client contacts https://prompeteer.ai/mcp and Prompeteer's OAuth discovery, registration, authorization and token endpoints on https://prompeteer.ai. Prompeteer sign-in may redirect to the identity provider selected by the user. Credentials remain in the client's supported OAuth flow; the plugin contains no tokens, API keys or environment-variable readers.
A Prompeteer account is required. Tool availability depends on the authenticated account, approved scopes and plan allowances. PromptDrive tools read or save account records; Memory tools access the user's Prompeteer Memory when enabled and authorized. Memory refers to Prompeteer's stored sources, not the host assistant's memory or conversation history. Generation and scoring can consume quota, and successful generation automatically saves its result to PromptDrive as part of the service workflow. The actual returned savedToVault and promptId fields indicate the persistence outcome. Generated skills are returned as content; this plugin does not execute them.
Privacy policy
Tool inputs and any requested saved-content or Memory results are processed by the Prompeteer service and returned to the connected client. Prompeteer's backend may use AI providers to carry out generation or scoring. See the published privacy policy for service data practices, terms, and plans. Contact info@prompeteer.com for support or privacy questions. Disconnect the MCP connection in the client when it is no longer needed.
Build and verify from source
Use Node.js 22.14.0 or later. The source lock pins the maintained proxy and its dependencies. Build copies those unmodified runtime files, with their licenses, into the release package.
npm ci --ignore-scripts
npm run build
npm run check
npm test
npm audit
npm pack --dry-run --json --ignore-scripts
The explicit build uses mcp-remote 0.8.6 and patched qs 6.16.0. Tests run protocol forwarding and startup-disconnect checks against local fixtures; they do not generate prompts or make paid service calls. The vendor/ build output is included in npm releases and is not committed to this repository.
Support and license
The bridge wrapper and documentation in this repository are MIT licensed. The hosted Prompeteer platform is proprietary.
常见问题
io.github.prompeteer/prompeteer-mcp 是什么?
Generate expert AI prompts for 140+ platforms with 16-dimension quality scoring.
相关 Skills
网页构建器
by anthropics
面向复杂 claude.ai HTML artifact 开发,快速初始化 React + Tailwind CSS + shadcn/ui 项目并打包为单文件 HTML,适合需要状态管理、路由或多组件交互的页面。
✎ 在 claude.ai 里做复杂网页 Artifact 很省心,多组件、状态和路由都能顺手搭起来,React、Tailwind 与 shadcn/ui 组合效率高、成品也更精致。
网页应用测试
by anthropics
用 Playwright 为本地 Web 应用编写自动化测试,支持启动开发服务器、校验前端交互、排查 UI 异常、抓取截图与浏览器日志,适合调试动态页面和回归验证。
✎ 借助 Playwright 一站式验证本地 Web 应用前端功能,调 UI 时还能同步查看日志和截图,定位问题更快。
前端设计
by anthropics
面向组件、页面、海报和 Web 应用开发,按鲜明视觉方向生成可直接落地的前端代码与高质感 UI,适合做 landing page、Dashboard 或美化现有界面,避开千篇一律的 AI 审美。
✎ 想把页面做得既能上线又有设计感,就用前端设计:组件到整站都能产出,难得的是能避开千篇一律的 AI 味。
相关 MCP Server
GitHub
编辑精选by GitHub
GitHub 是 MCP 官方参考服务器,让 Claude 直接读写你的代码仓库和 Issues。
✎ 这个参考服务器解决了开发者想让 AI 安全访问 GitHub 数据的问题,适合需要自动化代码审查或 Issue 管理的团队。但注意它只是参考实现,生产环境得自己加固安全。
Context7 文档查询
编辑精选by Context7
Context7 是实时拉取最新文档和代码示例的智能助手,让你告别过时资料。
✎ 它能解决开发者查找文档时信息滞后的问题,特别适合快速上手新库或跟进更新。不过,依赖外部源可能导致偶尔的数据延迟,建议结合官方文档使用。
by tldraw
tldraw 是让 AI 助手直接在无限画布上绘图和协作的 MCP 服务器。
✎ 这解决了 AI 只能输出文本、无法视觉化协作的痛点——想象让 Claude 帮你画流程图或白板讨论。最适合需要快速原型设计或头脑风暴的开发者。不过,目前它只是个基础连接器,你得自己搭建画布应用才能发挥全部潜力。