Cork
What it does
Keeps AI coding agents on one codebase from overwriting each other, and briefs each on its code.
- Connects 在云端运行
- Account or key Sign in to the service when you connect (OAuth)
- Category 开发者工具
安装
选择你使用的 AI 应用,然后复制设置。
通过 3/5 项检查 · 你的提问会经过开发者的服务器(mcp.corkcli.com) · ✓ 命令已检查 2026年11月14日
第一步:你用哪个 AI?
选择你用的应用,步骤会显示在下方。
可连接 Claude · 约 2 分钟
不需要密钥,不用安装程序 · 在 claude.ai 或桌面应用设置一次,即可在聊天中使用 · 免费账户可添加 1 个 · Team/Enterprise 方案由所有者(Owner)添加
打开 Claude 的 Connectors 页面
CustomizeConnectors
点击“Add” → “Custom” → “Web”,填写这 2 项
把这里的内容复制到 Claude 中同名的输入框,然后点击 Continue
Name · 名称CorkRemote MCP server URL · 地址https://mcp.corkcli.com在 Claude 检测到的登录设置页点击 Continue · Authentication 按作者说明选择(Sign in now、Sign in when needed 或 No sign in),然后点击 Add
在聊天中开启
+ConnectorsCork
试一试
用 Cork 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude 说正在使用 Cork,或回答中出现 Cork 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
能否连接 ChatGPT 取决于方案
完整 MCP:网页版 ChatGPT Business、Enterprise 或 Edu · Pro:以读取/获取权限连接,修改数据的工具可能用不了 · Plus:官方文档尚未确认,如果在插件页看到添加自定义 MCP 的选项可以试试 · 仅限网页
在网页版 ChatGPT 打开添加 MCP 的页面
Plugins+Add custom MCP server
填写名称和地址
输入名称,在 Connection 下粘贴地址,勾选风险提示,然后点击 Create as a plugin
Name · 名称CorkConnection · 地址https://mcp.corkcli.com在聊天中选择 Cork
开启新聊天,输入前先从插件列表中选择 Cork
试一试
用 Cork 告诉我你能帮我做什么,然后做一件给我看✓ 如果 ChatGPT 说正在使用 Cork,或回答中出现 Cork 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 OpenAI 帮助页面核实 · 你的界面可能因版本略有不同
可连接 Claude Code · 约 2 分钟
不需要密钥,不用安装程序
打开终端
Mac:按 Command+空格,输入 Terminal · Windows:打开 PowerShell
粘贴这条命令并按 Enter
claude mcp add --transport http --scope user corkcli https://mcp.corkcli.com在 Claude Code 中输入 /mcp
应能看到 Cork 在列表中并显示已连接
试一试
用 Cork 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Claude Code 说正在使用 Cork,或回答中出现 Cork 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同
可连接 Cursor · 约 2 分钟
不需要密钥,不用安装程序
在 Cursor 中以 Agent 模式打开 AI 聊天
试一试
用 Cork 告诉我你能帮我做什么,然后做一件给我看✓ 如果 Cursor 说正在使用 Cork,或回答中出现 Cork 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 Cursor 帮助页面核实 · 你的界面可能因版本略有不同
可连接 VS Code · 约 2 分钟
不需要密钥,不用安装程序
在 VS Code 中以 Agent 模式打开 AI 聊天
试一试
用 Cork 告诉我你能帮我做什么,然后做一件给我看✓ 如果 VS Code 说正在使用 Cork,或回答中出现 Cork 的工具名称,就表示连接成功
菜单名称已于 2026年11月11日 对照 VS Code 帮助页面核实 · 你的界面可能因版本略有不同
其他支持 MCP 的应用(如 Windsurf、Zed、Gemini CLI)添加方式各不相同,请在应用列表中查看
通过网页连接的应用使用的地址
https://mcp.corkcli.comJSON 设置在下方“开发者专用”中
开发者专用:终端命令和 JSON 设置
通过网页连接
Claude Code
claude mcp add --transport http --scope user corkcli https://mcp.corkcli.comCursor (~/.cursor/mcp.json)
{
"mcpServers": {
"corkcli": {
"url": "https://mcp.corkcli.com"
}
}
}VS Code (mcp.json)
{
"servers": {
"corkcli": {
"type": "http",
"url": "https://mcp.corkcli.com"
}
}
}这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
一览
哪些应用可以使用它
依据这个服务器的连接方式和各应用自己的文档,检查于 2026年11月10日。
通过互联网连接,适用于
ClaudeChatGPTGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatOpen WebUIRaycast
它能做什么
服务器要求先登录,才会列出它的工具。
基础检查i
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月14日
- ! 注意
作者与仓库一致
发布者(corkcli.com)不是仓库所有者(lithometric);可能是其他人代为上架的
- ✓ 通过
仍在维护
代码最后更新于37天前
- ✕ 未通过
有许可证
仓库中没有找到许可证文件
- ✓ 通过
端点有响应
在 164 毫秒内响应(HTTP 401)
- ✓ 通过
使用 OAuth 登录
声明支持 OAuth 登录,你的 AI 应用可以带你去登录
- … 尚未检查
工具描述中没有隐藏指令
尚未检查;我们会在检查端点时读取工具列表
讨论
提问、分享技巧或报告无法使用。按“有帮助”数量排序。
用户评价
还没有评价
与“开发者工具”中的其他服务器比较i
- 星标数多于此分类中其他 5871 个服务器的 0%
- 发布到注册中心的时间比此分类中其他 5871 个服务器的 99% 更新
- 基础检查结果等于或优于此分类中其他 5871 个服务器的 10%
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
常见问题
Cork 安全吗?
它在 5 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
它需要哪些密钥或代码?
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在我的电脑上运行还是在云端?
它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。
它还在维护吗?
GitHub 仓库最后更新于37天前,最新版本(1.0.0)于今天进入注册中心。
哪些应用可以使用它?
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,因此也能用于接受远程 MCP 的应用,例如网页版 ChatGPT(Business 及以上为完整 MCP;Pro 为只读)。
有什么可以替代它?
同一分类中有:REA, GitHub, Serena MCP: the IDE for your agent
同类别的其他选择
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
GitHub
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
Serena MCP: the IDE for your agent
A powerful toolkit for coding, providing semantic retrieval and editing capabilities.
作者的 README
打开 README
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Collide MCP: run multiple AI coding agents on one repo without conflicts
Collide is the coordination layer for multiplayer coding agents: it keeps Claude Code, Codex and Cursor agents, on one machine or across a team's machines, from overwriting each other on one codebase, and hands each agent what the others already learned.
This repository is Collide Free: run as many coding agents as you like on one machine, in as many repositories as you like, without them tripping over each other. Free, no account, and nothing leaves your machine.
curl -fsSL https://collidemcp.com/install.sh | sh
or, with Node (any terminal, including Windows PowerShell):
npx collidemcp
One install covers every repository on the machine. Agents already open pick it up at their next prompt.
Why
Run two or three Claude Code agents (or Codex, or Cursor) on the same codebase and they start colliding: one renames a function another is still calling, two write the same module, a push fails and the agent burns turns pulling and re-testing. Each agent also re-reads the same files the others just read. Separate git worktrees keep their files apart, but the work still meets at merge time.
Collide sits under the agents you already run. It does not run agents or host machines. Its hooks tell each agent what the others are changing while they work, hold a write that would clash, and hand each prompt the exact code it is about, so agents search less.
In Collide's benchmark (12 Claude Code agents at once on one repository, hard tasks), agents with Collide used 53% fewer tokens than the same agents without it. The study · all benchmarks
What your agents get
- A briefing before they start. The functions a prompt is about, with their signatures, who calls them and what changed recently, handed to the agent before it searches.
- Each other's work as it happens. When another agent renames a function yours uses, or writes the module your task needs, your agent hears about it on its next step.
- Writes that would clash are held. An edit against code another agent just changed waits until your agent has their version.
- Lint from teammates' changes. A rename or a new signature becomes a rule your agents are checked against.
- One-step pushes.
git pushbrings in the other agents' commits, runs your tests on the result and pushes, instead of the reject, pull, retry loop. - Messages between agents.
collide message "..."reaches the other agents in the repository, or one of them (--to), or whoever is on a piece of code (--about path.py::symbol). - Notes that stay with the code. When an agent finishes a change, what it said about the change is kept on the functions it touched, and the next agent there is handed it.
- MCP tools.
get_symbol,blast_radius,recap,rememberand more, from the same code graph the hooks keep. - What it saved. Once a day your agent's terminal shows the tokens Collide saved on this machine.
How it compares
| Collide | Git worktrees | Agent orchestrators and cloud agents | |
|---|---|---|---|
| Where agents run | Where they already run | Where they already run | On their servers or VMs |
| Agents see each other's changes while working | Yes | No, only at merge | Varies |
| Holds a write that would clash | Yes | No | No |
| Briefs each agent on the code its prompt is about | Yes | No | Varies |
| Works with Claude Code, Codex and Cursor together | Yes | Yes | Varies |
They combine: agents in separate worktrees still benefit from Collide telling them what the others are changing. Detailed comparisons: collidemcp.com/blog/compare.
Commands
| Command | What it does |
|---|---|
collide status | What Collide did on this machine, and where its data is |
collide message "..." | Tell the other agents in this repository something |
collide agents-md | Add Collide's section to this repository's AGENTS.md, after what is already there |
collide login | Link this machine to a Collide account |
collide upgrade | Bring in your team: start Team for your workspace |
collide uninstall | Take Collide's hooks out (your local data stays) |
Where your data is
Everything stays in ~/.collide/local on your machine: a small server on 127.0.0.1 that only this machine can reach. Collide Free makes no calls out, except to check for a signed update of itself.
Team
Collide Free connects the agents on one machine. Collide Team ($10 per seat per month, 14 days free) connects machines: your teammates' agents see your agents' work, and prompts are matched to code by meaning, not only by the names they use. collide upgrade moves a machine over with nothing else to download, and brings its history along.
Frequently asked questions
How do I run multiple Claude Code agents on the same repo without conflicts? Install Collide (one command above), then start your agents as usual. Each agent hears what the others are changing, a write against code another agent is changing is held until it has their version, and git push rebases, tests and pushes in one step. Guide: run multiple Claude Code agents on one repo.
Does it work with Codex and Cursor? Yes. The install registers Collide's hooks for Claude Code, Codex and Cursor, and its MCP server for any MCP client.
Is Collide an MCP server? It is both hooks and an MCP server. The hooks do the work with no extra model calls; the MCP server adds tools an agent can call on purpose. The hosted server is https://mcp.collidemcp.com.
Does my code leave my machine? Not on Collide Free. On Team, the structure of your code (names, signatures, who calls what) reaches Collide's servers; it is parsed on your machine and file contents are not sent.
Is it really free? Collide Free is free with no account and no time limit, for one machine.
Build from source
cd free && cargo build --release
./target/release/collide install
free/ builds the one program, collide: the hooks your agents run (hooks/, with the parser in core/) and the local server (collide local).
What is in this repository
| Folder | What it is | License |
|---|---|---|
hooks/ | What runs in your agent sessions: briefings, the write gate, landing, messages, the MCP bridge, install | MIT |
core/ | The code parser and graph the hooks and the engine share | MIT |
free/ | The engine: the local server that keeps the map, the briefings, lint, landing and memory | FSL-1.1-MIT |
plugin/ | The Claude Code plugin | MIT |
The engine's license lets you use, change and share it for anything except offering a competing product or service; each version becomes MIT two years after its release.
If Collide saved your agents a merge conflict, a star helps other developers running parallel agents find it. Questions and ideas: open an issue. Website: collidemcp.com · Claude Code plugin: lithometric/collidemcp