Aide 3168de4f

MCP运维云端

提供方 CrowdHailer

运维云端 未标注 无需鉴权 社区 / 第三方

用于构建符合MCP协议的远程服务器,支持JSON-RPC请求处理,适用于工具集成和资源访问场景。

一句话选型结论

按资料字段整理

Aide 3168de4f 属于运维云端类 MCP 服务器,传输方式未标注,无需鉴权,没有现成安装命令,需按官方文档接入。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类运维云端 传输方式未标注 鉴权无需鉴权 部署方式未标注
GitHub 星标 4 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

运维云端
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

适用场景

Gleam本地部署JSON-RPC服务器构建协议实现工具集成资源访问

安装与配置

未标注 无需鉴权
上游未给出传输方式
  1. 1
    准备运行环境

    按官方文档确认运行环境(常见为 Node.js 或 Python),装好后再执行安装命令。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
云资源查询

查实例、存储、账单等云上资源状态。

部署与发布

把部署、重启、回滚这类操作接进对话。

监控与告警

拉取监控指标,异常时让模型帮你判断。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 资料标注无需鉴权,装好就能连。

先想清楚的地方

  • 没有现成安装命令,需要按官方文档自己填启动命令和参数。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在运维云端场景里让模型直接调用外部能力的人。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Aide 3168de4f当前 未标注 未标注 未标注 无 2026-05-17
lobehub 未标注 stdio · 本地进程 本地部署 有 2026-04-11
TrendRadar 未标注 stdio · 本地进程 本地部署 有 2026-04-13
Ui Tars Desktop 8a825770 未标注 stdio · 本地进程 本地部署 有 2026-05-17
ruflo 未标注 stdio · 本地进程 本地部署 有 2026-04-13
FastGPT 未标注 stdio · 本地进程 本地部署 有 2026-04-11

服务器档案

运维云端
字段来自项目仓库与公开索引,中文说明由本站标注
提供方CrowdHailer上游标注的维护方 CrowdHailer
分类运维云端本站按上游分类展示
连接方式未标注上游未给出传输方式
鉴权方式无需鉴权远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · 未标注本地安装时用到的包标识与运行时
部署类型未标注上游给出的部署形态
最近核验2026-05-17上游最后核验时间

详细介绍

7.5 KB
来自项目仓库与上游目录的原文,未做删改

aide

Build Model Context Protocol (MCP) Servers (clients coming soon).

Install```sh

gleam add aide


## Usage

### Build a Server

This guide explains building an MCP server with some simple maths tools. See examples/math_server for the completed code.

Aide only supports building remote servers, not a problem as these can be used locally by directing your client to `localhost:`. From this point when we say MCP Server we implicitly mean Remote Server.

#### 1. Mounting an MCP server

An MCP server is a a single endpoint that accepts a JSON-RPC request and returns a JSON-RPC response.
This endpoint can be hosted on it's own or as part of a larger application.
Any choice of web server is valid.
This guide assume a wisp server.

A common approah to routing in wisp is to have a router module that pattern matches on path and method.
The route serving mcp is `/mcp`, not it needs to handle all methods.```gleam
// math_server/www/router
import gleam/http
import gleam/http/request.{Request}
import math_server/www/mcp
import wisp

pub fn handle(request, context) {
  use <- wisp.log_request(request)
  let Request(method:, ..) = request
  case wisp.path_segments(request), method {
    [], http.Get -> wisp.html_response("Check out our MCP server", 200)

    ["mcp"], _any -> mcp.handle(request, context)
    _, _ -> wisp.not_found()
  }
}

2. Handle JSON RPC

The MCP protocol builds on JSON RPC, again the aide library lets you choose your own approach to JSON decoding.

The handle function in the math_server/www/mcp module encapsulates decoding to MCP specific types, including decoding your applications specific tool types.

The handling of the mcp requests is another handle function defined in the module math_server/mcp.
Separating math_server/www/mcp and math_server/mcp allows for the same server to be used with other transports.```gleam
// math_server/www/mcp
import aide
import aide/definitions
import gleam/dynamic/decode
import gleam/http
import gleam/http/request.{Request}
import gleam/json
import gleam/option.{Some}
import gleam/string
import math_server/mcp
import wisp

pub fn handle(request, context) {
let Request(method:, ..) = request
case method {
http.Delete -> wisp.no_content()
http.Post -> {
// decode input to a MCP request
use mcp_request <- decode_json(request, aide.request_decoder())

  // create an mcp server config.
  // In this example the server is the same for all users, but if different tools are required for different users,
  // this is where you would create a user specific server.
  let server = server(context)

  // handle mcp request in the mcp module
  mcp.handle(mcp_request, server)
  // encode the response and return it
  |&gt; option.map(aide.response_encode)
  |&gt; option.map(json.to_string)
  |&gt; option.map(wisp.json_response(_, 200))
  |&gt; option.unwrap(wisp.response(202))
}
_ -&gt; wisp.method_not_allowed([http.Post, http.Delete])

}
}

fn decode_json(request, decoder, then) {
use data <- wisp.require_json(request)
case decode.run(data, decoder) {
Ok(value) -> then(value)
Error(reason) -> wisp.bad_request(string.inspect(reason))
}
}

fn server(_context) {
aide.Server(
implementation: definitions.Implementation(
name: "math_server",
version: "0.1.0",
title: Some("Math Server"),
),
tools: mcp.tools(),
resources: [],
resource_templates: [],
prompts: [],
)
}


This server has the same tools for all users.

### 3. Implement server functionality.

Now we get to the business end of our MCP server.

Aide chooses an continuation passing style to implement the server functionality.
*This makes supporting JavaScript and BEAM runtimes easier.*
The details don&#x27;t matter here as our example doesn&#x27;t require promise or result types.

In the this snippet `call_tool`, `read_resource`, `get_prompt` and `complete` are all functions we need to implement.
As this server only implements tools we can create noop versions of all functions but `call_tool`.```gleam
// math_server/mcp
import aide
import aide/effect
import aide/tool
import gleam/dict
import gleam/dynamic/decode
import gleam/int
import gleam/result
import oas/generator/utils
import oas/json_schema

pub fn handle(mcp_request, server) {
  case aide.handle_rpc(mcp_request, server) {
    effect.Done(result) -&gt; result
    effect.CallTool(tool:, resume:) -&gt; resume(call_tool(tool))
    effect.ReadResource(resource:, resume:) -&gt; resume(read_resource(resource))
    effect.GetPrompt(prompt:, resume:) -&gt; resume(get_prompt(prompt))
    effect.Complete(ref:, argument:, context:, resume:) -&gt;
      resume(complete(ref, argument, context))
  }
}

// continued in Implement tools

4. Implement tools

We need to implement the tools we want to expose to the client.
A tool in MCP has a specification and a decoder to decode the input from the client.
The specification is used to generate the OpenAPI schema for the tool, that the LLM will understand.

Check out aide_generate to create encoders/decoders from the tool input/output schemas.

MCP requires that input and output are objects.
There for the input/output schemas expect a list of json_schema fields, so you can't accidentally specify an invalid schema.```gleam
// ...
// math_server/mcp

pub type Tool {
Random
Add(Int, Int)
}

pub fn tools() {
[
tool.Tool(
spec: tool.Spec(
name: "random",
title: "Generate Random",
description: "Generate a random number between two numbers",
input: [],
output: [json_schema.field("number", json_schema.integer())],
),
decoder: decode.success(Random),
),
tool.Tool(
spec: tool.Spec(
name: "add",
title: "Add",
description: "Add two numbers",
input: [
json_schema.field("x", json_schema.integer()),
json_schema.field("y", json_schema.integer()),
],
output: [json_schema.field("sum", json_schema.integer())],
),
decoder: {
use x <- decode.field("x", decode.int)
use y <- decode.field("y", decode.int)
decode.success(Add(x, y))
},
),
]
}

fn call_tool(tool) {
case tool {
Random -> {
use number <- result.map(math.random())
dict.from_list([#("number", utils.Integer(number))])
}
Add(x, y) -> {
use number <- result.map(math.add(x, y))
dict.from_list([#("number", utils.Integer(number))])
}
}
}


## Development```sh
gleam test

MCP definitions

The module aide/definitions is generated from a JSON Schema specification, maintained by the MCP project.
To run the generation run.gleam run dev

OAS doesn't support list of types.

  • type is changed to true on additionalProperties on ElicitResult
  • RequestId is true
  • ProgressToken is true

The Request type is not referenced anywhere in the definitions, but is used to specify a JSON RPC request
Most other XRequest definitions are a contant and parameters

MCP

Model context protocol should allow for easy access from chats to tools and resources.
This doesn't seem to be the case so far.

Claude

Seems to ignore structured_content field in response and only use conent

Mistral

Connectors are limited to Gmail and Google Calendar for Free and Pro.
There is no mention of custom connectors, although no details are shared about the enterprise level.

OpenAI

Needs a Pro or Team Plan

https://aaronparecki.com/articles

Credit

Created for EYG, a new integration focused programming language.

常见问题

5 问
答案由本站依据条目字段整理,不引用上游原文
Aide 3168de4f 怎么安装?

资料没有给出安装命令,请到官方仓库或文档按说明接入;这类条目多数是生态应用,或尚未发布可安装包。

Aide 3168de4f 怎么在客户端里配置?

它的传输方式未标注,在客户端 MCP 配置里按官方文档填写启动命令和参数,保存后重启客户端生效。

Aide 3168de4f 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Aide 3168de4f 需要 API Key 吗?

资料标注为无需认证,装好就能连;如果实际调用时报鉴权错误,按官方文档补上凭据。

Aide 3168de4f 能提供哪些能力?

资料显示能力标签包括 服务器构建、协议实现、工具集成。具体工具清单以官方文档为准。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-05-17。