Skip to content

webkubor/image-agent-plus

image-agent-plus

One sentence → great image. No API key needed.

Smart AI image generator powered by Codex CLI or Gemini CLI —
auto prompt expansion, reusable profiles, zero configuration.

Changelog · Report Bug · npm

npm version License Stars


AI 时代,我们应该沉淀下来的是什么


What it does

npm install -g image-agent-plus
image-agent-plus generate --prompt "a tabby cat on a rainy windowsill"

That's it. No API key. No config. Just install Codex CLI, and you're generating.

Key features:

  • 🚫 No API key required — uses your local Codex CLI login
  • 🧠 Smart prompt expansion — short prompts auto-enriched from your profile
  • 🎯 Reusable profiles — set once, generate forever (social / blog / wallpaper / poster …)
  • 🤖 Agent-native — skills for Codex, OpenClaw, and Hermes included
  • 🔀 Multi-provider — Codex · Gemini · OpenAI, switchable per call

Quick start

# 1. Install Codex CLI (one-time setup, no API key needed)
npm install -g @openai/codex
codex login

# 2. Install image-agent-plus
npm install -g image-agent-plus

# 3. Generate
image-agent-plus generate --prompt "neon Tokyo street at night, rain"

Profiles — set once, skip forever

Profiles save your default platform, aspect ratio, and style so you never repeat yourself.

# See all profiles (* = active default)
image-agent-plus profile list

# Switch default
image-agent-plus profile use story      # 9:16  stories/reels
image-agent-plus profile use social     # 3:4   instagram / 小红书
image-agent-plus profile use blog       # 16:9  web blog
image-agent-plus profile use wallpaper  # 16:9  desktop
image-agent-plus profile use avatar     # 1:1   profile photo

# Create your own
image-agent-plus profile set \
  --name youtube \
  --platform youtube-thumbnail \
  --aspect-ratio 16:9 \
  --style "bold colorful"

Built-in profiles:

Name Platform Ratio Style
social instagram/xiaohongshu 3:4 photorealistic
story stories/reels 9:16 photorealistic
blog web-blog 16:9 modern illustration
hero web-hero 21:9 cinematic
wallpaper desktop-wallpaper 16:9 artistic
avatar profile-photo 1:1 photorealistic
poster print-poster 3:4 editorial

Profiles saved to ~/.image-agent-plus/profiles.json.


Smart prompt expansion

Short prompts are automatically enriched when a profile is active:

Input:   "一只橘猫"
Profile: social (3:4, photorealistic, instagram/xiaohongshu)

Output:  "一只橘猫, portrait composition, subject in upper-center,
          lifestyle aesthetic, warm tones, soft bokeh, shareable,
          photorealistic, high detail, professional quality, sharp focus,
          subject sharp and detailed, soft ambient light, minimal noise"

Pass --no-expand to use the raw prompt as-is.


Generate

# Uses active profile — aspect ratio, style, provider auto-applied
image-agent-plus generate --prompt "cherry blossoms at dusk"

# Override aspect ratio for this call
image-agent-plus generate --prompt "wide city skyline" --aspect-ratio 16:9

# Use a specific profile for this call
image-agent-plus generate --prompt "product shot" --profile avatar

# Multiple images
image-agent-plus generate --prompt "abstract art" --output-count 4

Runtime check

image-agent-plus check
State Default provider
Codex CLI installed codex — no API key ever needed
Only Gemini CLI gemini — needs GEMINI_API_KEY for image output
Neither installed guided install message

Override per call: --provider codex|gemini|openai

Optional env vars for direct API mode:

export IMAGE_AGENT_GEMINI_API_KEY="your_key"
export IMAGE_AGENT_OPENAI_API_KEY="your_key"

Agent skills

Skill What it does
image-prompt-refiner Checks profile, asks only missing questions, expands prompt
reference-style-transfer Extracts style from a reference image and builds a transfer prompt
image-agent-plus Direct CLI invocation skill

Agent commands (.toml): smart · profile · generate · image-agent


CLI reference

Flag Default Description
--prompt required Image description
--profile active default Profile to apply
--no-expand false Skip prompt expansion
--aspect-ratio from profile 16:9 9:16 1:1 3:4 21:9
--provider auto codex gemini openai
--output-count 1 1–8 images
--filename ~/Desktop/image-agent-plus-output/ Output path
--seed Fix seed for reproducibility
--preview / --no-preview Open after generation

Development

git clone https://github.com/webkubor/image-agent-plus
cd image-agent-plus
pnpm install && pnpm run build
node bin/image-agent-plus.js check

License

Apache License 2.0. Fork lineage includes Google's original nanobanana.

About

A Gemini CLI extension for per-call Nano Banana 2 / Pro switching, aspect ratio control, and image generation without server restarts.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors