Put the project you are working on online without leaving your editor. Install once, sign in once in the browser, then say "put this online". Your AI tool deploys the current folder (or a GitHub repo) to Dockhold and reports a live HTTPS URL. It can also read status and logs, set variables, and resize apps and managed databases.
No token to copy. Sign-in happens through the browser, the same way as the , and every client below shares it.
Deploy your project to a live HTTPS URL from your AI tool; read logs, set variables, resize apps.
Every install below starts the same local MCP server, npx -y dockhold mcp. It forwards each request to Dockhold with the sign-in the CLI stored on your machine. Nothing sensitive goes into your editor config.
This installs the MCP server and a deploy skill (invoked as /dockhold:deploy, or picked up on its own when you ask to put the project online) that knows the whole flow: preflight, sign-in, deploy, and what to do when a build fails.
This repo is also an Agent Plugin (plugin.json and mcp.json at the root), so Cursor can install it from its plugin browser once it is listed there: open Customize in the sidebar, search for Dockhold, and choose Install.
This repo is also a Gemini CLI extension (gemini-extension.json at the root), so the install adds the MCP server and the same deploy skill.
Cline
Ask Cline to "install the MCP server from https://github.com/dockhold/mcp". It follows llms-install.md: one settings entry, no token, then a browser sign-in. To do it by hand, add the entry below to cline_mcp_settings.json with "env": { "DOCKHOLD_REF": "cline" }.
Any other MCP client
Use the same server entry in your client's config:
If your client gives a server only a few seconds to start, the first npx -y dockhold download can be too slow for it. Install the CLI once with npm i -g dockhold and the same entry starts instantly.
Sign in
The first time a tool needs your account it answers with a message that says to run npx dockhold login in a terminal. Run it, a browser tab opens, and you are done. The sign-in lasts 90 days and is shared by the CLI and every client above. Nothing else to configure.
What it looks like
You: put this online
AI: Checked the project: it listens on PORT and has a Dockerfile. You are not signed in yet, so I am running npx dockhold login. A browser tab will open.
(you sign in, the tab says you are done)
AI: Signed in. Deploying with npx dockhold deploy --name recipe-box. Uploading, building...
AI:(calls get_app_logs) The last 100 lines: the server started on port 8080 and has served 3 requests. No errors.
Local folders deploy through the CLI, which needs no GitHub account. A GitHub repository deploys through the deploy_app tool. Deploys on every push need the repository connected to Dockhold (list_github_repos gives the github_installation_id to pass); a repository deployed by URL alone builds once.
Tools
Tool
Scope
What it does
list_apps
read
List your apps with id, status, and URL
get_app_status
read
Deploy status, URL, last commit, error message for one app
get_app_logs
read
Recent runtime logs for one app
get_resource_usage
read
Account pools (compute, database RAM, storage) and per-app sizes
list_github_repos
read
Your connected GitHub repos, with the id needed to deploy private ones
deploy_app
deploy
Deploy a GitHub repo as a live web app
deploy_group
deploy
Deploy several connected services in one go
redeploy_app
deploy
Zero-downtime restart of an app
set_app_variable
deploy
Set a non-secret environment variable and roll it out
unset_app_variable
deploy
Remove a plain environment variable and roll it out
resize_app
deploy
Change an app's memory
resize_database
deploy
Change a managed database's memory
resize_database_storage
deploy
Grow a managed database's disk
set_app_storage
deploy
Give an app storage that survives restarts and deploys, or grow it
remove_app_storage
deploy
Remove an app's storage and erase its files
The browser sign-in has the read and deploy scopes, which covers every tool above.
Remote server (no CLI)
For a client that cannot run a local command, the same tools are served over Streamable HTTP at https://api.dockhold.eu/mcp. Create a token in the Dockhold dashboard under Settings, then API tokens (read-only, read+deploy, or with secrets), and pass it as a bearer:
Claude Code: claude mcp add --transport http dockhold https://api.dockhold.eu/mcp --header "Authorization: Bearer dh_mcp_...". This is the one setup that puts a token in a config file, so keep that file out of version control.
A dashboard token can also carry the secrets scope, which unlocks five more tools over this connection: set_app_secret, unset_app_secret, bind_app_secret, list_secrets, list_app_secrets. The browser sign-in never has that scope; with the plugin, secrets are added in the dashboard's Secrets section.
Getting started with Dockhold
Dockhold is the easiest way to put your app online: push from GitHub or deploy from a folder, always on, secure by default. Sign in at app.dockhold.eu. The free tier needs no card.
Noch keine Tool-Liste. Remote-Server erhalten eine nach dem ersten Health-Check.
FAQ
Fragen
Wie installiere ich Dockhold?
Kopieren Sie auf dieser Seite die http-Konfiguration für Cursor, Claude Desktop, Claude Code oder VS Code. Sie nutzt die Installation des Herausgebers, keinen PlugMCP-Proxy.
Benötigt Dockhold einen API-Schlüssel?
Der Herausgeber nennt diese Umgebungsvariablen: DOCKHOLD_REF.
Welchen Transport verwendet Dockhold?
Dockhold wird über HTTP erreicht.
Wann wurde Dockhold zuletzt geprüft?
Zuletzt geprüft: 2026-11-08. Deploy your project to a live HTTPS URL from your AI tool; read logs, set variables, resize apps.