GitHub
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
作者提供的说明(英文原文)
Dokploy MCP server: manage apps, Docker Compose, databases, domains, backups, and servers.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
DOKPLOY_URL, DOKPLOY_API_KEY, TRANSPORT_TYPE, PORT, HOST标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"dokploy": {
"command": "npx",
"args": [
"-y",
"dokploy-mcp-server"
],
"env": {
"DOKPLOY_URL": "your-value",
"DOKPLOY_API_KEY": "your-value",
"TRANSPORT_TYPE": "your-value",
"PORT": "your-value",
"HOST": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add dokploy --env DOKPLOY_URL=your-value --env DOKPLOY_API_KEY=your-value --env TRANSPORT_TYPE=your-value --env PORT=your-value --env HOST=your-value -- npx -y dokploy-mcp-server运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"dokploy": {
"command": "npx",
"args": [
"-y",
"dokploy-mcp-server"
],
"env": {
"DOKPLOY_URL": "your-value",
"DOKPLOY_API_KEY": "your-value",
"TRANSPORT_TYPE": "your-value",
"PORT": "your-value",
"HOST": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"dokploy": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"dokploy-mcp-server"
],
"env": {
"DOKPLOY_URL": "your-value",
"DOKPLOY_API_KEY": "your-value",
"TRANSPORT_TYPE": "your-value",
"PORT": "your-value",
"HOST": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
TRANSPORT_TYPE envPORT envHOST env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。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
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7日
注册名称已通过 GitHub 账户 sapientsai 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
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。
它在 1 项适用的基础检查中通过了 1 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。
你需要填写:DOKPLOY_URL, DOKPLOY_API_KEY
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(1.10.0)于34天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
同一分类中有:GitHub, Desktop Commander, Cloudflare MCP
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ 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
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
MCP server for terminal commands, file operations, and process management
Cloudflare MCP servers
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。
A comprehensive Model Context Protocol (MCP) server for Dokploy - the open-source, self-hosted PaaS. Deploy apps, manage containers, databases, domains, and servers through AI assistants like Claude.
Tracks Dokploy v0.30.8. Tools are built against the v0.30.8 OpenAPI spec (604 paths). The comparison below was measured on 2026-10-03.
The official Dokploy MCP generates one tool per API endpoint. Its latest release, @dokploy/mcp@0.30.7, exposes 604 tools, one for each of the 604 paths in Dokploy's v0.30.8 OpenAPI spec. Coverage is complete, and every one of those schemas loads into the model's context before you ask your first question.
This server hand-curates the same API into 27 tools (one per category, each taking an action enum), covering the deploy-and-operate surface most self-hosters use daily.
Official @dokploy/mcp | This server | |
|---|---|---|
| Tools exposed | 604 | 28 (27 + info) |
tools/list schema payload | 319,143 bytes | 51,974 bytes |
| Approximate tokens loaded up front | ~80k | ~13k |
| Median tool schema | 387 bytes | 1,257 bytes |
Measured 2026-10-03 against @dokploy/mcp@0.30.7 (the latest published) and this server's build tracking Dokploy v0.30.8. Each server was started over stdio, sent tools/list, and its serialized tool array counted in bytes (divided by 4 for a rough token estimate).
On a 200k-token context window, the official server spends about 40% of it before you ask anything; this server spends about 6.5%. The larger median schema here is deliberate: descriptions carry the workflow knowledge that prevents failed calls, such as which service id pairs with which databaseType.
Tool counts for the official server are per category, taken from its live tools/list at @dokploy/mcp@0.30.7.
| Category | Official MCP | This Server |
|---|---|---|
| Projects | 11 tools | 1 tool (6 actions) |
| Applications | 32 tools | 1 tool (24 actions) |
| Compose | 31 tools | 1 tool (21 actions) |
| Deployments | 9 tools | 1 tool (5 actions) |
| Docker | 21 tools (incl. disk usage) | 1 tool (17 actions) |
| Docker Volumes | 8 tools | 1 tool (8 actions) |
| Docker Images | 3 tools | 1 tool (3 actions) |
| Networks | 9 tools | 1 tool (9 actions) |
| Overview | 3 tools | 1 tool (3 actions) |
| DNS Providers | 11 tools | 1 tool (11 actions) |
| Vault Providers | 7 tools | 1 tool (7 actions) |
| Domains | 10 tools | 1 tool (9 actions) |
| Redirects | 4 tools | 1 tool (4 actions) |
| Servers | 19 tools | 1 tool (9 actions) |
| Settings | 52 tools | 1 tool (5 actions) |
| Databases | 94 tools (6 engines) | 1 tool (17 actions, all 6 DB) |
| Backups | 12 tools | 1 tool (6 actions) |
| Volume Backups | 6 tools | 1 tool (6 actions) |
| Preview Deployments | 4 tools | 1 tool (4 actions) |
| Schedules | 6 tools | 1 tool (6 actions) |
| Audit Log | 1 tool | 1 tool (1 action) |
| Environments | 7 tools | 1 tool (6 actions) |
| Infrastructure | 13 tools (ports, certs, auth) | 1 tool (8 actions) |
| Mounts | 6 tools | 1 tool (6 actions) |
| SSH Keys | 7 tools | 1 tool (6 actions) |
| Registries | 7 tools | 1 tool (7 actions) |
| Destinations | 6 tools | 1 tool (6 actions) |
| Total | 399 tools | 27 tools |
Key advantages:
dbType + action paramsaction enum parameter; other params are optional based on actionOf the official server's 604 tools, 399 fall inside the categories above; the remaining 205 have no equivalent here:
If your workflow needs any of those, use the official server, or run both.
Add to your MCP configuration:
{
"mcpServers": { "dokploy": { "command": "npx", "args": ["-y", "dokploy-mcp-server"], "env": { "DOKPLOY_URL": "https://dokploy.example.com", "DOKPLOY_API_KEY": "your-api-key" } } } }
Add to .cursor/mcp.json:
{
"mcpServers": { "dokploy": { "command": "npx", "args": ["-y", "dokploy-mcp-server"], "env": { "DOKPLOY_URL": "https://dokploy.example.com", "DOKPLOY_API_KEY": "your-api-key" } } } }
docker run -e DOKPLOY_URL=https://dokploy.example.com \
-e DOKPLOY_API_KEY=your-api-key \ -e TRANSPORT_TYPE=httpStream \ -p 3000:3000 \ dokploy-mcp-server
| Variable | Required | Default | Description |
|---|---|---|---|
DOKPLOY_URL | Yes | - | Your Dokploy instance URL |
DOKPLOY_API_KEY | Yes | - | API key from Dokploy Settings > API Keys |
TRANSPORT_TYPE | No | stdio | Transport mode: stdio or httpStream |
PORT | No | 3000 | HTTP port (httpStream mode only) |
HOST | No | 0.0.0.0 | HTTP host (httpStream mode only) |
Each tool uses an action enum to select the operation. Parameters are optional and used based on the chosen action.
dokploy_project (6 actions)Actions: list | get | create | update | remove | duplicate
Manage projects. list and get return nested environments with their applications, composes, and databases (with names, IDs, and status), so you can discover service IDs without extra calls. create requires name. update requires projectId + fields. remove requires projectId. duplicate requires sourceEnvironmentId + name.
Reading the status label. Dokploy's status enum is idle | running | done | error, rendered as [IDLE], [RUNNING], [DONE], [ERROR]. It describes the deployment, not live container state — [DONE] means the last deploy finished, and [IDLE] means nothing has run, neither of which implies a container is up right now. Check dokploy_docker getContainers for what is actually running.
dokploy_application (24 actions)Actions: create | get | update | move | deploy | start | stop | delete | markRunning | refreshToken | cleanQueues | killBuild | cancelDeployment | reload | saveEnvironment | setEnvVars | getEnvKeys | getEnvValuesUnsafe | saveBuildType | traefikConfig | readMonitoring | readLogs | search | deployNginxQuickstart
Full application lifecycle. Most actions require applicationId. create requires name + environmentId. deploy supports redeploy flag. readMonitoring requires appName. search finds applications by q/name/appName/repository/owner/dockerImage/projectId/environmentId, with limit (1–100, default 20) and offset; the reply flags a truncated page. deployNginxQuickstart takes environmentId (+ optional serverId) and creates and deploys a "Hello World" nginx demo app on a generated traefik.me domain — Dokploy's onboarding smoke test.
Env handling. get returns a masked env summary (count only) — never the values. Three actions cover the rest:
setEnvVars — granular merge inside the server. Pass set (KEY=VALUE per line, upsert) and/or unset (array of KEY names). The read-modify-write happens server-side; the tool result is a masked confirmation listing only the changed key names. No untouched-key values ever enter the transcript.getEnvKeys — returns KEY names only, sorted. Safe to read.getEnvValuesUnsafe — unsafe escape hatch that returns the full KEY=VALUE blob. Use only when you genuinely need the values; the output is in the tool transcript and any retained agent logs.saveEnvironment — full-replace, kept for explicit blob-set workflows.sourceType:
github → repository + owner + branch (+ githubId for private)git → customGitUrl + customGitBranchdocker → dockerImagebuildType: dockerfile | heroku_buildpacks | paketo_buildpacks | nixpacks | static | railpack.
The underlying API also supports gitlab/bitbucket/gitea/drop sources, but those need provider-specific fields not yet exposed by this tool.
dokploy_compose (21 actions)Actions: create | get | update | delete | deploy | start | stop | move | loadServices | loadMounts | getDefaultCommand | cancelDeployment | cleanQueues | killBuild | refreshToken | saveEnvironment | setEnvVars | getEnvKeys | getEnvValuesUnsafe | readLogs | search
Docker Compose management. Most actions require composeId. create requires name + environmentId. deploy takes optional redeploy and freshVolumes — `freshVolumes` runs `docker compose down --volumes` first, deleting the service's volume data (docker-compose type only). loadMounts requires serviceName. cancelDeployment/cleanQueues/killBuild/refreshToken require composeId. search takes the same query fields as dokploy_application.
Env handling. Same shape as dokploy_application: get returns a masked summary, setEnvVars merges, getEnvKeys lists key names, getEnvValuesUnsafe is the escape hatch, saveEnvironment full-replaces.
sourceType:
github → repository + owner + branch (+ composePath)git → customGitUrl + customGitBranch (+ customGitSSHKeyId for private)-
只显示了开头部分;其余内容请在仓库中阅读