draw-cli

Generate images from text prompts via the Hugging Face Inference API — designed to be called from any shell, script, or AI coding agent without leaving the terminal.

Why agents use this

A coding agent can generate placeholder art, hero images, mock UI assets, or concept sketches inline from its shell session, without switching context or calling a web API manually.

# Generate a placeholder hero image for a landing page being built
draw "minimalist SaaS dashboard hero, dark theme, 16:9" -o assets/hero.png

# Produce an icon concept during component work
draw "flat vector icon, a glowing terminal cursor, transparent background" -o src/icons/cursor.png

# Pipe a dynamically assembled prompt from another tool
echo "isometric 3D render of a microservice architecture diagram, pastel colors" | draw -o docs/arch.png

The output is a plain image file — drop it straight into the asset pipeline, send it to Figma, or attach it to a Telegram report via tg --photo.

Install

One-liner (installs deps, links draw into PATH, registers the agent skill):

curl -fsSL https://github.com/alex-mextner/draw-cli/main/install.sh | bash

Alternative — isolated env via pipx:

pipx install git+https://github.com/alex-mextner/draw-cli

Either way, finish with token setup (see below), then optionally run the skill registration step manually if you skipped the one-liner:

draw install-skill

install-skill is idempotent — it writes a skill file to ~/.agents/skills/draw/ so Claude Code, Codex, opencode, and Gemini harnesses know draw exists. The one-liner runs it automatically.

Token setup

Create ~/.config/draw-cli/.env:

HF_TOKEN=hf_...

Get a token at huggingface.co/settings/tokens (a free-tier read token is enough for inference).

Usage

# Prompt as positional arg
draw "a cute robot" -o robot.png

# Override model
draw "a cute robot" --model black-forest-labs/FLUX.1-dev -o robot.png

# Prompt from stdin
echo "a cute robot" | draw -o robot.png

Flags

Flag	Default	Description
prompt (positional)	—	Text prompt. Reads from stdin if omitted.
-o / --out	required	Output image path (e.g. out.png).
--model <hf-id>	black-forest-labs/FLUX.1-schnell	Any HF text-to-image model ID.

Env vars

Variable	Description
HF_TOKEN	Hugging Face access token. Auto-loaded from ~/.config/draw-cli/.env.
HF_MODEL	Default model override (same effect as --model).

Requirements

• Python 3.9+
• huggingface_hub and Pillow Python packages

---

How draw compares

The other text-to-image CLIs trade off between simple-but-locked-in and powerful-but-heavy. Single-vendor tools (dallecli, openai-cli-art) are one pip install but hard-wired to OpenAI and a paid key. Model runners (Replicate CLI, simonw llm) are flexible but route through a paid hosted API or a general LLM harness. Local engines (comfy-cli / ComfyUI) are the most capable but pull in a full generative stack and a server.

draw is the minimal middle: one command, any Hugging Face text-to-image model via --model (FLUX by default), runnable on a free-tier HF token, stdin-pipeable for agent-assembled prompts, and it registers an agent skill so harnesses discover it. It deliberately does one thing — prompt in, image file out — and leaves editing/filtering to real image tools.

Tool	Model-agnostic	Free-tier path	Stdin pipe	No local server	Agent-skill registration	Single-purpose simplicity
draw	✓ (any HF model)	✓ (HF free token)	✓	✓	✓	✓
dallecli	— (OpenAI only)	— (paid key)	—	✓	—	~ (also edit/filter)
openai-cli-art	— (OpenAI only)	— (paid key)	~	✓	—	~
Replicate CLI	✓ (any hosted model)	~ (free credits)	✓	✓	—	— (generic runner)
simonw llm (+ image plugins)	✓ (via plugins)	~ (depends on backend)	✓	~	—	— (general LLM CLI)
comfy-cli / ComfyUI	✓ (local + partner)	✓ (local)	—	— (runs a server)	—	— (full stack)

~ = partial. draw is not the most powerful — comfy-cli wins on local control and llm on breadth — but it is the lightest path from a shell prompt to an image file with no vendor lock-in and no server to babysit, which is exactly what a coding agent needs for placeholder and concept art.

Ecosystem

Part of the HyperIDE.ai agent toolchain:

• tg-cli — simple Telegram CLI to send messages, photos & files, and a two-way agent bridge (reports, Q→buttons, voice/rich)
• review-cli — multi-model read-only code review from one command: diff review, cited quorum, brainstorm, visual review, and interactive spec-review tooling. Read-only, CLI-first, harness-agnostic.
• rig-cli — umbrella dev-env driver: sets up a repo from config — skills, hooks, CI, dep-bootstrap; reconciles drift
• agent-tools — the shared catalog rig applies: portable agent skills, agent-hooks, the global git-hook dispatcher, CI gates, and MCP servers
• 3d-cli — scriptable CLI for the full 3D FDM lifecycle: modeling, mesh repair, slicing, and print monitoring
• hyperide.ai — Figma replacement inside VS Code. Edit React components directly through AST/LSP without AI hallucinations, token waste, or context-window limits. Works for indie vibe-coding and for enterprise teams with split design/dev roles.

Each CLI registers a skill into your agent harnesses (<tool> install-skill) so agents know it exists — see Install.