Use the Judt CLI

The Judt CLI manages Apps from the command line. Each command is one call to the Judt API, and its help names that call. Use the CLI when you have a shell but no MCP client.

Install the CLI

To install the CLI, use one of these commands:

Method	Command
Script (macOS, Linux)	`curl -fsSL https://judt.ai/install \| sh`
Go	`go install github.com/judt-ai/cli/cmd/judt@latest`

All methods install the same binary. To see how to update your copy, run judt upgrade. Use one of these methods:

judt auth login                  # Opens your browser
judt auth login --device         # Gives a code to enter on another device
judt auth login --with-token < key-file

For CI jobs and agents, set JUDT_API_KEY. This variable has priority over a saved sign-in:

read -rs JUDT_API_KEY && export JUDT_API_KEY

To check your sign-in, run judt whoami. To learn which credential to use, see Authentication.

Manage an App

Commands under judt apps act on one App. To set the App for a directory, run judt apps use. The CLI saves the App in .judt.json. judt apps create saves the new App in the same way. To set the App for one command, use --app.

judt apps create --name "Acme hiring"
judt apps files write src/index.ts --content @src/index.ts
judt apps run "npm test"
judt apps preview
judt apps commit "Add the pricing page"
judt apps deploy
judt apps logs --level error

To set a Secret, send the value on stdin, so that it is not in your shell history:

pbpaste | judt apps secrets set STRIPE_SECRET_KEY

To call an API operation directly, use judt api:

judt api GET /api/apps

All groups and commands have --help. The help tells what the command does, which App it acts on, its flags, and the API call that it makes.

Use the CLI in a script

Add --json to get the JSON response of the API.
Lists return one page. Add --all to get all pages, or --cursor with the next_cursor of the previous page.
Text flags take @path to read a file and - to read stdin.
If a write fails with 503 or 504, stderr shows an --idempotency-key value. Run the same command again with that flag to get the result of the first run. Without the flag, the work runs again.

The exit code tells you the result. judt apps run exits with the exit code of your command.

Exit code	Meaning
`0`	The command succeeded.
`1`	The API refused the request, or the request failed. On an error, stderr shows `code: message`.
`2`	The command was not used correctly.
`4`	You are not signed in, or you do not have permission.

Environment variables

Variable	Description
`JUDT_API_KEY`	The API key to use in place of the saved sign-in.
`JUDT_APP`	The App to act on.
`JUDT_PROFILE`	The saved sign-in to use. You can also use `--profile`.
`JUDT_API_URL`	Another API, such as staging.
`JUDT_CONFIG_DIR`	The directory for sign-ins and settings. The default is `~/.config/judt`.
`JUDT_NO_KEYRING`	Stores sign-ins in a file, not in the OS keychain.
`JUDT_NO_UPDATE_CHECK`	Stops the check for new versions.
`JUDT_NO_MODIFY_PATH`	Stops the install script from changing your shell’s startup file.

Next steps

See all API operations in the API reference.
Connect an MCP client in Connect to MCP.
Limit an agent to one App in Give an agent access to a single App.

Get started

Concepts

Guides

Features

Reference

Use the Judt CLI

Install the CLI

Manage an App

Use the CLI in a script

Environment variables

Next steps

Install the CLI

Sign in

Manage an App

Use the CLI in a script

Environment variables

Next steps