layer

Command-line interface for Layer — AI asset generation for game studios.

Reaches the same models, workspaces, and Creative Unit balance as the app, from a terminal, a Makefile, or a CI job.

Install

uv tool install layer        # or: pipx install layer

Python 3.10 and up — deliberately older than the services that build it, so the same package imports inside Blender, Maya, and Houdini.

Working on the CLI itself? Install it from your checkout instead, so the layer command follows the branch you have out:

task -d apps/cli-py install-cli

Sign in

layer login                 # opens a browser
layer login --device        # for SSH sessions and containers
layer whoami
layer use my-studio

For CI and other unattended runs, set LAYER_API_KEY to a personal access token instead of signing in.

Generate

layer generate image -p "a crystal sword icon, transparent background"
layer generate video -d 5 --aspect-ratio 16:9 -p "camera pushes through the gate"
layer generate mesh  -p "weathered treasure chest, game-ready"
layer generate audio -p "arcade coin pickup"

Each waits for the run, prints what it cost, and writes the result into the working directory under a name derived from the prompt — never overwriting a file already there. -o takes a directory, or the exact filename when there is a single output; --no-download prints URLs instead.

Leave --base-model off and the recommended model for that modality is chosen and named. The catalogue is read from your workspace, not shipped with the package:

layer base-models list --modality video
layer base-models show bfl-flux-2-pro      # the keys --param accepts

Every run is priced before it executes. --confirm asks first; --max-cost 8 refuses anything more expensive. For a long render, submit and collect later:

id=$(layer --json generate video --no-wait -p "..." | jq -r .inference_id)
layer generate status "$id" --wait

Where settings come from

Highest wins:

1. a command-line flag
2. an environment variable — LAYER_PROFILE, LAYER_WORKSPACE, LAYER_PROJECT, LAYER_API_URL, LAYER_API_KEY
3. layer.toml, found by walking up from the working directory
4. the active profile in ~/.layer/config.toml

Committing a layer.toml pins a checkout to its workspace, so two repos open in two terminals cannot generate into each other's:

workspace = "my-studio"
project   = "season-12-art"

Scripting

--json puts the result on stdout and nothing else — progress, warnings, and errors all go to stderr, so a pipeline never has to filter them out.

layer --json whoami | jq -r .workspace
layer --json generate image -p "..." | jq -r '.files[].path'

Exit codes: 0 ok, 1 error, 2 usage, 3 auth, 4 insufficient balance, 5 invalid input, 6 content policy, 7 timeout. They may gain members, never change meaning, so a caller can branch on them without parsing prose.

Documentation

https://www.layer.ai/docs/cli

Driving Layer from an agent rather than a shell? There is an MCP server and an Agent Skills pack.

Licence

Apache-2.0