rifts.to MCP server
Enables creating, managing, and analyzing live audience surveys on rifts.to directly from an AI client. Supports creating surveys, listing them, retrieving results, and closing surveys via natural language.
README
rifts.to MCP server
An MCP server for rifts.to, a live audience survey tool. It lets an AI client create a survey, list the surveys on your account, read back the results, and close a survey when you're done collecting responses.
Point a poll at your audience by asking your AI client to do it, and check the answers the same way, without leaving the chat.
claude mcp add --transport http rifts https://mcp.rifts.to/mcp
That is the hosted server, and it signs you in through the browser. Running your own copy is supported too.
Tools
create_survey: creates a survey from a title and a list of questions and returns its public link and admin link. Optionally accepts a custom slug.list_surveys: lists the surveys on your account, with title, status, and response count. Admin links are included only if you ask for them, so a routine listing doesn't hand a model a pile of credentials it didn't need.get_survey_results: returns a survey's questions and every response.close_survey: stops a survey from accepting new responses. Reopening, retitling, and editing questions stay in the rifts.to admin dashboard, not here.
Hosted, with sign-in
rifts.to runs this server at https://mcp.rifts.to/mcp. Add it as a remote MCP
server and your client walks an OAuth flow: it registers itself, sends you to
rifts.to to approve, and stores the token it gets back. Nothing to paste, and
you can revoke it later under "Connected applications" on /account.
In Claude Desktop, add it as a custom connector with that same URL.
Running it yourself, with a token you paste in, is the rest of this document.
Run it yourself
npx @rifts_to/mcp
Releases are published from CI with npm provenance, so each version on npm carries a signed attestation linking it to the commit and workflow run that built it. You can check the one you installed:
npm audit signatures
Cutting a GitHub release triggers the publish; the tag has to match the version
in package.json or the run fails.
Building from source works too, and is the same code:
git clone https://github.com/riftsto/mcp.git rifts-mcp
cd rifts-mcp
npm install
npm run build
node dist/stdio.js
The server reads its configuration from two environment variables:
RIFTS_TOKEN(required): your personal access token.RIFTS_API_URL(optional): defaults tohttps://rifts.to. Set this only if you're pointing the server at a different rifts.to deployment.
Claude Desktop
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"rifts": {
"command": "npx",
"args": ["-y", "@rifts_to/mcp"],
"env": {
"RIFTS_TOKEN": "your-token-here"
}
}
}
}
Running from a clone instead means "command": "node" with
"args": ["/path/to/rifts-mcp/dist/stdio.js"].
Claude Code
claude mcp add rifts -e RIFTS_TOKEN=your-token-here -- npx -y @rifts_to/mcp
Add -s user to make it available in every project instead of just the
current one.
Getting a token
- Sign in at rifts.to.
- Go to
/account. - Create a token under "API tokens."
Token creation requires an active subscription. The token is shown once, at creation time, so save it somewhere before you close the dialog. If you lose it, create a new one.
Self-hosting
This server is stateless and holds no credentials of its own. Every request carries your RIFTS_TOKEN and goes straight to the rifts.to API; the server doesn't store it, cache it, or log it anywhere. Running your own copy, on your own machine or your own infrastructure, is a supported way to use it, and it still needs a token from a real rifts.to account with an active subscription. There's no separate self-hosting tier or discount; the token is what's paid for, not the server.
Development
npm install
npm test # vitest, drives the server over an in-memory transport
npm run typecheck
npm run build
The tests connect a real MCP client to the server and call the tools against a stub API, so a schema that rejects a valid question, or a result the SDK refuses to serialize, fails in CI rather than in someone's chat window.
src/server.ts builds the server and is transport-free. src/stdio.ts and
src/worker.ts are the two entry points, and neither contains tool logic, so
the hosted and self-hosted paths cannot drift on what the tools do.
Security
Token handling, what this server can reach, and how to report a vulnerability: SECURITY.md.
License
Apache-2.0. See LICENSE.
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。