CLI

The andesine CLI works with the Andesine API from your terminal and your scripts. It needs Node.js 22.12 or later, and is open source under the MIT license.

npm install --save-dev andesine @andesine/sdk
npx andesine --help

Use npm install --global andesine to install it for your whole system.

Sign in

npx andesine auth login

The CLI opens Andesine in your browser and shows a code. Check that the code on the Authorize CLI page is the same, and select Approve. Use --no-browser to open the address yourself.

The CLI stores your sign-in in the credential store of your system: Keychain on macOS, Credential Manager on Windows, or Secret Service on Linux. If none is available, it can store it in a file, after you confirm.

In CI, set ANDESINE_API_KEY instead. The API key is used before a saved sign-in.

Use profiles for more than one account or instance, for example andesine --profile work auth login.

Set up a project

npx andesine init

init asks for the workspace, the content source (published or current), the collections, and the path of the generated types. It saves them in andesine.json:

{
  "version": 1,
  "workspaceID": "ws_...",
  "types": {
    "source": { "kind": "published", "channel": "published" },
    "collections": ["/Docs"],
    "output": "src/andesine.generated.ts"
  }
}

Commands use the nearest andesine.json. The file has no credentials, so you can commit it. Use init --mcp to also set up the local MCP server.

Flags and environment variables override the file:

Call the API

andesine api <resource> <operation> runs any operation of the API reference. Names use kebab case: content.getTree is api content get-tree.

npx andesine api content get-tree --channel published
npx andesine api content get --slug-path /docs/introduction --channel published
npx andesine api content list-entries --channel published --paginate
npx andesine api search ask-published-stream --input @question.json --format text
npx andesine api assets upload --asset-id ast_... --file ./image.png

• --help shows the fields of an operation, and --schema shows its input JSON Schema.

• Pass fields as flags, such as --collection-id, or as JSON with --input @file.json or --input - for stdin.

• Output is one JSON value per line. --full adds the HTTP status and headers. --paginate prints each page.

• --output <file> saves binary responses, such as images.

Other commands

Output and exit codes

Data goes to stdout, and messages and prompts go to stderr, so you can pipe the output to other tools. Prompts are off in CI and with --no-interactive.