Crescendo Lab
繁體中文Book a demo

Agent Lab CLI · Crescendo Agent Lab Deploy

Push a Docker image
and deploy it.

The Agent Lab CLI signs you in and lets docker push to your organization's container registry. Push the image you built, then deploy it as a new version of a project.

Jump to the commandsStart free

Commands

Install, sign in, push, deploy

Replace <org-id> with your organization's ID. The console's Image registry page shows these commands with it filled in.

  1. 1

    Install the CLI

    npm install --global @crescendolab/agentlab-cli@alpha
agentlab --version

    Install the CLI from npm. It needs Node.js 24 or later and runs on macOS, Linux and Windows, on x64 and arm64. It is an alpha release, so commands may still change.

  2. 2

    Sign in and set up Docker

    agentlab login
agentlab auth configure-docker
agentlab status

    agentlab login opens your browser to sign in and choose an organization. agentlab auth configure-docker then makes docker use that sign-in for the registry, with no docker login.

  3. 3

    Build and push the image

    docker build --platform linux/amd64 -t registry.agentlab.cresclab.com/<org-id>/web:v1 .
docker push registry.agentlab.cresclab.com/<org-id>/web:v1

    Build for linux/amd64, the only platform projects run on, and push it under your organization's ID.

  4. 4

    Deploy it

    On the project's Deploy tab, choose An image under Deploy a new version from, then pick the repository and tag. A project's first version goes live once it is ready; after that, select Go live.

    Building from source instead? See Deploy from a Dockerfile.

Image contract

What the image must do

The same rules apply to an image you push and to one built from a Dockerfile.

RUNTIME

Port, user, platform

  • Serve HTTP on $PORT: the port in the project's settings, else the one port the image exposes, else 8080.
  • An image with no non-root USER runs as uid 10001, group 0.
  • linux/amd64 only; a push for another platform is refused.
  • A CMD or ENTRYPOINT starts the server.
  • The filesystem is read-only apart from /tmp.
CLI

Commands

  • agentlab login [--org <uuid>] signs in through your browser.
  • agentlab auth configure-docker makes docker use the sign-in for the registry.
  • agentlab status shows the account, organization and token expiry.
  • agentlab logout signs out and removes this machine's saved sign-in.

Every rule and notice: Container runtime contract. CLI install details: the package on npm.

CLI or MCP

Two ways in, one platform

The MCP server is for deploying from an AI tool. The CLI is for Docker images you build yourself.

Scroll the table sideways.

The Agent Lab CLI compared with the Agent Lab MCP server
Agent Lab CLIAgent Lab MCP server
Use it whenYou already build a Docker image, by hand or in CIYou ask Claude Code, Codex or Cursor to deploy
What it sendsA built image, pushed with dockerSource files, built on the platform
Sign-inagentlab login in the browserOAuth in the browser, from the AI tool
Installnpm package (alpha)One command per AI tool; nothing to install

FAQ

Before you push

How do I install the Agent Lab CLI?
Run npm install --global @crescendolab/agentlab-cli@alpha. It needs Node.js 24 or later and works on macOS, Linux and Windows, on x64 and arm64. Go is not required.
Do I need docker login?
No. Run agentlab login, then agentlab auth configure-docker once, and docker pushes as you. For CI, create a robot token in the console's deploy settings and use it with docker login.
What address do I push to?
registry.agentlab.cresclab.com/<org-id>/<name>:<tag>, where <org-id> is your organization's ID. The console's Image registry page shows the full command.
Can I push an arm64 image from an Apple silicon Mac?
Projects run on linux/amd64 only, and a push for another platform is refused. Build with docker build --platform linux/amd64.
Which plans include the container registry?
Pro, Team and Business; the Free plan doesn't include it. The plans page lists how many image versions and how much storage each plan keeps. Plans and pricing
How is the CLI different from the MCP server?
The MCP server lets an AI tool such as Claude Code or Codex deploy your source for you. The CLI is for when you build the Docker image yourself and push it. Set up the MCP server