CLI
Generate images, video and music from the terminal, scripts and AI agents
geckogen is the GeckoGen command-line client. It uses the same models, prices, balance and generation history as the website.
Install
curl -fsSL https://api.geckogen.app/cli/install.sh | sh
The installer downloads one self-contained binary for your platform and checks the signed release manifest before installing it. Node.js is not needed.
Sign in
geckogen auth login
Sign-in opens in the browser and is confirmed by your GeckoGen session — there is no key to copy. On a machine without a browser, use geckogen auth login --manual.
Pick a model and read its parameters
Every model has its own parameters. List the models, then read the parameter schema of the one you picked:
geckogen list_models
geckogen get_model google/nano-banana-pro
Generate
geckogen generate google/nano-banana-pro \
'{"prompt":"app icon, a jar with a scan frame"}' \
--wait --output-dir ./out
The server validates the parameters against the model's schema and answers with a clear error if a value is wrong. --wait blocks until the generation finishes, and --output-dir downloads the result.
Useful commands
| Command | What it does |
|---|---|
geckogen quote_generation | Shows the token price of a generation before you start it |
geckogen get_me | Shows your token balance |
geckogen list_generations | Lists your generation history |
geckogen wait_generation <id> | Waits for a started generation to finish |
geckogen artifact download <id> <dir> | Downloads every output of a finished generation |
geckogen batch | Runs a resumable JSONL manifest of generations |
geckogen update | Updates the CLI to the latest release |
Run geckogen <command> --help for the flags of any command.
A generation that fails returns its tokens to your balance. Stopping the CLI does not stop a generation that has already started.