闪耀 YMCP 构建 MCP R中的应用。一个MCP App 是一个带有页面的工具：一个支持 MCP 应用的聊天客户端 扩展，如Claude、ChatGPT或VS Code，会在 对话，这个人可以继续使用它。在闪光 工具是R函数，页面是Shiny UI。模型可以调用 工具本身，在任何MCP客户端中，页面调用相同的工具 随着使用者使用。

安装

一个工具中的应用

名称连接了两半。工具的参数 bins 取其值 从带有该ID的输入开始，返回的列表填充输出 histogram 和 caption。当用户移动滑块时，页面 再次呼叫工具。

模型调用了同一个工具，并读取它返回的内容，无论是否 其客户端可以显示该页面：

图也以图像的形式达到它。每个结果分为三种： 模型的文本和紧凑的结构化数据，以及页面绘制的内容 （HTML、图片、控件数据），客户端传递给页面并保留这些内容 从模型中获取。结果可以包含文本、表格、图表、图片、HTML和 htmlwidgets。

preview_app(app) 在你的浏览器中打开应用的方式，就像聊天客户端一样

shinymcp

shinymcp builds MCP Apps in R. An MCP App is a tool with a page: a chat client that supports the MCP Apps extension, such as Claude, ChatGPT, or VS Code, shows the page in the conversation, and the person can keep working with it. In shinymcp the tools are R functions and the page is Shiny UI. The model can call the tools on their own, in any MCP client, and the page calls the same tools as the person uses it.

Installation

# install.packages("pak")
pak::pak("JamesHWade/shinymcp")

An app from a tool

library(shiny)
library(shinymcp)

faithful_histogram <- ellmer::tool(
  function(bins = 20) {
    list(
      histogram = mcp_result_plot(
        function() hist(faithful$waiting, breaks = bins, main = NULL, xlab = "Minutes"),
        text = "Histogram of waiting times between eruptions."
      ),
      caption = sprintf("%d eruptions; median wait %d minutes.", nrow(faithful), median(faithful$waiting))
    )
  },
  name = "faithful_histogram",
  description = "Show a histogram of waiting times between Old Faithful eruptions.",
  arguments = list(
    bins = ellmer::type_integer("Number of bins, 5 to 50.", required = FALSE)
  ),
  annotations = ellmer::tool_annotations(read_only_hint = TRUE)
)

app <- mcp_app(
  ui = fluidPage(
    sliderInput("bins", "Number of bins", min = 5, max = 50, value = 20),
    plotOutput("histogram"),
    textOutput("caption")
  ),
  tools = list(faithful_histogram),
  name = "faithful"
)

Names connect the two halves. The tool’s argument bins takes its value from the input with that id, and the list it returns fills the outputs histogram and caption. When the person moves the slider, the page calls the tool again.

The model calls the same tool, and reads what it returns, whether or not its client can show the page:

result <- app$run_tool("faithful_histogram", list(bins = 30))
cat(result$content[[1]]$text)
#> histogram: Histogram of waiting times between eruptions.
#> 
#> caption: 272 eruptions; median wait 76 minutes.