Deploy from your terminal or your CI pipeline
The MultiConnect CLI builds, uploads and rolls out your server, API or website with one command, and checks on any deployment databases included. It runs on Windows, macOS and Linux, drops straight into GitHub Actions, GitLab CI or any pipeline, and lets your AI assistant deploy and report for you.
Quick Start
Get running in three commands
# END CODE
Already installed? Update to the latest version with: dotnet tool update –global MultiConnect.
Features
Everything You Need in the Terminal
One-command deploys
Runs your build command, packages the output, uploads it and rolls it out then waits until it is live.
Zero-downtime rollouts
The current build keeps serving until every new container is healthy, and is restored automatically if the new one never becomes healthy.
Players stay connected
For game servers with a grace period, old servers keep running so connected players are not dropped and your pipeline does not wait for them.
Change settings as you deploy
Use deploy –config to update environment variables, start command, instance counts, autoscaling or on-demand game server limits, and website options. The new settings and the new build go live together.
Status at a glance
Check state, build version, connection details, usage, cost per hour and your credit balance for servers, APIs, websites and databases.
Built for CI/CD
No prompts in CI, clear exit codes, and keys read from pipeline secrets.
Works with your AI assistant
Add the MultiConnect skill to Claude, Cursor, GitHub Copilot or ChatGPT and ask it to deploy or check on your deployments. It asks for your approval before every command.
Secure by design: deploy keys
The CLI signs in with an access key that has “Allow deploys (CLI)” turned on not with your password.
- Create the key in the dashboard, then attach it to the deployments it may deploy.
- A deploy key can only reach the deployments it is attached to.
- It can upload builds, redeploy and read status, connection details, usage and your balance.
- It cannot create, delete or stop deployments or change plans those stay in the dashboard.
- Every request is signed with HMAC-SHA256, including the request body, so a captured request cannot be altered or replayed.
- Changed or replaced the key? Run multiconnect init --force in the project folder to store the new one.
- Deploys made with the CLI are recorded in your account's activity log, marked as done through the CLI and with which key.
- Deploy keys
Commands
Command Reference
Useful deploy options: –tag (label the release), –skip-build, –no-wait, –timeout <minutes>, –yes.
| Command | What it does |
|---|---|
| multiconnect init | Choose which deployment this folder is for and how to build it. Saves multiconnect.json |
| multiconnect deploy | Build, package, upload and roll out with the deployment's current settings, then wait until live |
| multiconnect deploy --config | The same, but first change settings interactively; settings and build go live together |
| multiconnect status | State, build version, any rollout in progress, and the containers |
| multiconnect info | Connection details |
| multiconnect balance | Credit balance, spend per hour and how long the balance will last |
| multiconnect analytics | Cost per hour, usage against the plan, and CPU / memory / storage per container |
Useful deploy options: –tag (label the release), –skip-build, –no-rollback (keep a failed rollout in place instead of restoring the previous build), –no-wait, –timeout <minutes>, –yes.
Databases have no build to deploy, so deploy does not apply to them; status, info, balance and analytics all work. Master-worker servers deploy their master and worker builds together. Container logs are available in the dashboard.
Let your AI assistant deploy for you
The MultiConnect skill teaches an AI assistant to use the CLI: deploy a new build, check status, usage and cost, and read your balance. Then just ask “deploy this to MultiConnect” or “how is my game server doing?”.
- AI assistants
Approval before every command
It asks for your approval before every command, including read-only ones, and explains exactly what each command will change.
Your key stays yours
It never asks for your key. Run
multiconnect init yourself, once, in your own terminal.
Focused on what the CLI supports
It redeploys and reports using the CLI, then points you to the dashboard for everything else.
Add it to your assistant
Choose your AI tool and follow the setup instructions.
Claude Code
Unzip the skill so the folder sits at
~/.claude/skills/multiconnect-cli/
for every project, or
.claude/skills/multiconnect-cli/
inside one project.
Claude apps
For Claude desktop and claude.ai, add a custom skill in Claude's settings and upload the ZIP file.
Cursor, Copilot & others
Add the contents of SKILL.md as project rules
or custom instructions in ChatGPT, Cursor, GitHub Copilot,
and other compatible assistants.
Before you begin: Your assistant needs a terminal where it can run commands, with the MultiConnect CLI installed. The same downloads are available on the CLI page of your dashboard.
CI/CD
Deploy from GitHub Actions
Commit multiconnect.json, store the key and secret as pipeline secrets, and run deploy.
Frequently asked questions
Everything you need to know about MultiConnect Cloud.
No. Create the deployment once in the dashboard, attach a deploy key, and use the CLI for every release after that.
Yes. It holds the linked deployment and your build settings. The key inside it is encrypted for the machine and user that ran init, so a committed copy is useless anywhere else. In CI, the key comes from pipeline secrets instead.
The current build keeps serving. If the new containers never become healthy, the previous build is restored automatically and the CLI exits with a non-zero code.
Yes. A key can be attached to several deployments, and each project folder is linked to exactly one of them.
Not yet. Live logs and log downloads are on each deployment’s Logs page in the dashboard.
No. You link the folder yourself with multiconnect init, and the key is stored encrypted on your machine. The skill tells the assistant never to ask for it, never to put it on a command line and never to print it.