Changesets MCP

MCP浏览器工具

提供方 未标注

浏览器工具 stdio · 本地进程 OAuth 授权 社区 / 第三方
查看接入方式

全面解析BlahMCP Server的核心功能、安装配置和实用案例。作为顶级Model Context Protocol服务器,Blah能让AI助手访问实时数据、执行操作,为您提供更智能的工作体验和自动化解决方案。

一句话选型结论

按资料字段整理

Changesets MCP 属于浏览器工具类 MCP 服务器,stdio · 本地进程,OAuth 授权,有可直接执行的安装命令。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。

分类浏览器工具 传输方式stdio · 本地进程 鉴权OAuth 授权 部署方式本地部署
GitHub 星标 未标注 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

浏览器工具
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

安装与配置

stdio · 本地进程 OAuth 授权
客户端在本机拉起子进程,数据不出本机
npx changeset add\",

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Node.js,再按安装命令拉取依赖。

  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
自动操作网页

打开页面、点击、填表,把浏览器动作交给模型执行。

抓取结构化数据

把页面内容整理成表格或 JSON,省掉手工复制粘贴。

关键流程回归

按步骤检查页面是否正常,回归时少漏环节。

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

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 使用前要准备好OAuth 授权对应的凭据,并按最小权限配置。
  • 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

  • 需要在浏览器工具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

  • 完全不想在本地装环境的人。
  • 不方便管理密钥或第三方授权的人。
  • 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
Changesets MCP当前 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Playwright MCP Server 21 stdio · 本地进程 本地部署 有 2026-05-18
Fetch 9a2682de 1 SSE · 服务端推送 支持远程部署 有 2026-05-18
Browser 738d52bc 未标注 stdio · 本地进程 本地部署 有 2026-05-17
MCP 40600ece 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Skyvern MCP 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

浏览器工具
字段来自项目仓库与公开索引,中文说明由本站标注
提供方未标注上游标注的维护方 未标注
分类浏览器工具本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式OAuth 授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 · Node.js本地安装时用到的包标识与运行时
部署类型本地部署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

9.8 KB
来自项目仓库与上游目录的原文,未做删改

PROJECT IN EXTREME POC MODE - NOTHING STABLE AT ALL - DOCS AND VISION ARE PURELY NOTES

BLAH - Barely Logical Agent Host - Update #1

(think npmjs for tools)

Once this list is fully done, the BLAH JSON schema will be finalized.

  • [ ] Tools
  • [x] Local functions
  • [x] Hosted functions
  • [ ] Registry
    • [ ] Publish
    • [ ] Browse
  • [x] Mcp support
    • [x] Local Stdio
    • [x] Local SSE
    • [ ] Hosted SSE
  • [x] Slop Support
    • [x] Local Slop
    • [x] Hosted Slop
  • [ ] Tool Composability
    • [ ] Grouping
    • [ ] Flows
    • [ ] Triggers / Tags
  • [ ] Extends
    • [x] Local blah.json
    • [ ] Hosted blah.json
  • [ ] Debugging
    • [ ] Playground
    • [ ] Logging

Notes:
- The flows will be using a subset of agnt.gg's flow schema
- Extend means you can compose blah.json's in any way you like. (think what if you just want to include a designers you follow tools)
- The registry means we will have a convenient way to just push and pull functions (tools)

BLAH - Barely Logical Agent Host

BLAH is an open-source ecosystem for managing, distributing, and executing AI agent tools using the Model Context Protocol (MCP). It provides a decentralized registry for MCP servers that promotes transparency, security, and community-driven development.

Project Structure

This project is organized as a monorepo using Turborepo:

  • packages/cli - The main BLAH CLI tool for working with MCP
  • apps/web - Web interface for BLAH
  • apps/docs - Documentation site

Getting Started

Prerequisites

  • Node.js >= 18.18.0 (Node 20+ recommended)
  • pnpm

Installation```bash

Clone the repository

git clone https://github.com/thomasdavis/blah.git
cd blah

Install dependencies

pnpm install

Build the project

pnpm run build


### Quick Start with the CLI

See the CLI README for detailed instructions on using the BLAH CLI.```bash
# Navigate to the CLI package
cd packages/cli

# Set up your environment variables (.env file)
echo "OPENAI_API_KEY=your_openai_api_key_here" > .env
echo "BLAH_HOST=https://ajax-blah.web.val.run" >> .env

# Run a simulation
pnpm run simulate

Currently working on

  • [ ] Need a way to debug mcp servers using any model/ide
  • [ ] Think about portals
  • [ ] lisa watts came up with "auton", likely what i am going to suggest

Unrelated Ideas

  • [ ] Need save state in ideas so you don't lose stuff (if you hadn't comitted)
  • [ ] upvotes for social tooling
  • [ ] automatic error rate
  • [ ] share your tool combination results

Using a local blah.json

Hosted blah.json (Valtown)

Getting started (hosted)

  1. Setup a Valtown account
  2. Create a new HTTP function named blah````typescript export default async function server(request: Request): Promise<Response> { const tools = [ { name: "hello_name", description:Says hello to the name, inputSchema: { type: "object", properties: { name: { type: "string", description:Name to say hello to`,
    },
    },
    },
    },
    ];

return new Response(JSON.stringify(tools), {
headers: {
"Content-Type": "application/json",
},
status: 200,
});
}


### Adding to your Client

- Claude
  - Desktop
  - Code (CLI)
- Cursor
- Cline
- Windsurf
- BLAH Client
- Are there any nice web clients yet

### Development

I am still working on tools to make development not shit.

### Logging

While developing this, I want to send logs from everybodies clients to just figure out all the fucking errors the clients are throwing due to different configures.

### Playground

So far just a basic client that lists prompts, resources, and tools. Could be interactive later, and should also run against tests.```bash
npm run playground

Needs improvement

  • [ ] Valtown is the current wrapper, need to be an alternative maybe using Vercel functions or some shit.
  • [ ] Need a nice strategy for composability

Credits

  • Lisa Watts
  • Travis
  • Wombat

AI GENERATED BULL SHIT (the above is authentic)

  • mcp server version badge
  • registry uptime status

What is BLAH?

BLAH is an open-source ecosystem for managing, distributing, and executing AI agent tools using the Model Context Protocol (MCP). It provides a decentralized registry for MCP servers that doesn't suffer from misaligned incentives, promoting transparency, security, and community-driven development.

Key Features:

  • Open-source infrastructure that any system (IDE, AI platform, etc.) can connect to
  • Language-agnostic tool registry supporting unlimited tool selection
  • Robust security through optional signing and verification of MCP servers
  • Comprehensive CLI for publishing, discovering, and managing tools
  • Support for various tool types: functions, REST endpoints, local files, or standard manifests

🔍 Vision

BLAH aims to be the foundation for a new generation of AI tools that can be easily shared, discovered, and composed. We envision a future where:

  1. Anyone can create and share tools that extend AI capabilities
  2. Everyone has access to a rich ecosystem of tools regardless of their technical background
  3. Every system can integrate with this ecosystem through standard protocols

🚀 Getting Started

Installation```bash

npm install -g blah-mcp


### Basic Usage```bash
# Search for tools
blah search "image generation"

# Install a tool
blah install awesome-image-generator

# List installed tools
blah list

# Get info about a tool
blah info awesome-image-generator

📖 Core Concepts

The Registry

BLAH's registry is designed with lessons learned from package managers like npm, focusing on:

  • Transparency: All infrastructure code is open-source
  • Decentralization: Storage options include local, cloud, IPFS, or gists
  • Security: Optional signing of MCP servers with verification
  • Governance: Community-driven with no single entity controlling the ecosystem

Agent Behavior

BLAH supports various agent interaction patterns:

  • Parallel execution
  • Recursive tool calls
  • Branching workflows
  • Sequential processing
  • Complex graph-based workflows

Tool Definition

A tool in BLAH is fundamentally a function, but can be represented as:

  • Code snippets
  • RESTful endpoints
  • Local executables
  • Standard manifests (SLOP, agents.json)

Every tool should explicitly document how it's invoked, though this isn't mandatory.

🛠️ CLI Commands

BLAH comes with a comprehensive CLI:```
blah publish - Publish a tool to the registry
blah search - Search for tools by name, tag, or description
blah install - Install a tool
blah remove - Remove an installed tool
blah update - Update tools to latest versions
blah list - List installed tools
blah info - Display information about a tool
blah config - Configure BLAH settings
blah login - Authenticate with registry
blah logout - End authentication session
blah whoami - Show current authenticated user
blah version - Display version information
blah help - Show help information


## 📋 Manifests

BLAH uses a manifest file (`blah.json`) to define tools and their dependencies. This manifest can exist:

- In a project workspace
- In the home directory
- Hosted in the cloud
- As a gist or on IPFS

Example `blah.json`:```json
{
  "name": "awesome-image-generator",
  "version": "1.0.0",
  "description": "Tool for generating awesome images",
  "entry": "./dist/index.js",
  "tools": [
    {
      "name": "generate_image",
      "description": "Generates an image from a text prompt",
      "parameters": {
        "prompt": {
          "type": "string",
          "description": "Text description of the desired image"
        },
        "style": {
          "type": "string",
          "enum": ["realistic", "cartoon", "abstract"],
          "default": "realistic"
        }
      }
    }
  ],
  "dependencies": {
    "image-gen-lib": "^2.0.0"
  },
  "tags": ["image", "generation", "creative"]
}

🔄 Standards Support

BLAH supports and aims to be compatible with:

  • Model Context Protocol (MCP)
  • SLOP (Structure for Language Operator Protocols)
  • agents.json
  • [Your own custom protocols through converters]

🌟 Discovery and Metadata

Tools in BLAH are organized using:

  • Tags (e.g., #TOOLBELT)
  • Semantic metadata for intuitive navigation
  • Usage analytics for popularity-based recommendations
  • User-based recommendations (tools used by similar users)

🏗️ Architecture

BLAH consists of three main components:

  1. Registry: For tool storage and discovery
  2. CLI: For interacting with the registry and managing tools
  3. MCP Server: For executing tools within agent workflows

Currently, ValTown serves as the primary backend for registry storage and compute, though alternative implementations are welcome.

🔒 Security & Accountability

BLAH takes security seriously:

  • Optional signing of MCP servers
  • Code execution in sandboxed environments
  • Comprehensive logging for accountability
  • Compliance with local laws and regulations

🚧 Roadmap

  • [ ] Host the MCP server on cloud platforms (e.g., Cloudflare Workers)
  • [ ] Create a web-based tool explorer and documentation site
  • [ ] Improve debugging and developer experience
  • [ ] Develop a tool creation wizard/generator
  • [ ] Implement tool recommendation system

🧠 Future Ideas

  • User behavior sharing for personalized tool recommendations
  • Scheduled agent execution through cron-like mechanisms
  • Decentralized governance model for registry management
  • Extension ecosystem for custom registry implementations

🤝 Contributing

Contributions are welcome! Check back soon for detailed contribution guidelines.

👏 Credits

Special thanks to Lisa and Wombat for their invaluable contributions to this project.

📄 License

BLAH is released under the MIT License. See the LICENSE file for details.


Build Logical Agents, Humanely

常见问题

7 问
答案由本站依据条目字段整理,不引用上游原文
Changesets MCP 怎么安装?

在客户端按下面的安装命令接入:npx changeset add\",。装完在客户端配置里引用同一个包或端点。

Changesets MCP 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。

Changesets MCP 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。

Changesets MCP 需要 API Key 吗?

资料标注需要OAuth 授权认证,使用前准备好对应凭据,并按最小权限配置。

Changesets MCP 能提供哪些能力?

资料显示能力标签包括 版本管理、JavaScript、构建工具、自动化。具体工具清单以官方文档为准。

Changesets MCP 需要什么运行环境?

运行时为 Node.js;属于本地进程,需要在本机装好运行环境。

Changesets MCP 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

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。