ai-gateway

A tiny CLI for generating text, images, and video via the Vercel AI Gateway. One key, hundreds of models, no code required.

ai-gateway "explain quicksort in 3 bullets"
ai-gateway image "a red fox in a snowy forest" -o fox.png
ai-gateway video "a wave crashing on rocks at sunset" --duration 5
cat README.md | ai-gateway "summarize this"

Install

Requires Node 20+.

npm install -g @vesely/ai-gateway-cli

Or with another package manager:

pnpm add -g @vesely/ai-gateway-cli
bun add -g @vesely/ai-gateway-cli

First run

Get an API key from https://vercel.com/ai-gateway and either:

• Run any command — you'll be prompted to paste it (saved to ~/.config/@vesely/ai-gateway-cli/config.json, chmod 600), or
• Export it: export AI_GATEWAY_API_KEY=..., or
• Run ai-gateway config set key sk_...

Commands

| Command | What it does | | --- | --- | | ai-gateway "<prompt>" | Streamed text completion (default command). | | ai-gateway image "<prompt>" | Generate an image, save to disk. | | ai-gateway video "<prompt>" | Generate a video, save to disk. Takes minutes. | | ai-gateway models | List all available models with prices. | | ai-gateway config | Show current config. | | ai-gateway config set <key> <value> | Set key, text-model, image-model, or video-model. |

Defaults

Picked for the best price/quality ratio:

• Text: xai/grok-4.1-fast-non-reasoning
• Image: google/imagen-4.0-fast-generate-001
• Video: xai/grok-imagine-video

Override per-call with -m <model-id>, or persist with ai-gateway config set text-model openai/gpt-5.4.

Examples

# Text with a different model
ai-gateway -m anthropic/claude-opus-4.6 "draft a tweet about CLIs"

# Pipe context in
git diff | ai-gateway "write a commit message for this diff"

# Get raw JSON (no streaming)
ai-gateway --json "hello" | jq .text

# Image with a specific output path
ai-gateway image -m bfl/flux-2-pro -o cover.png "a minimalist mountain logo"

# Multiple images
ai-gateway image -n 4 "abstract wallpaper, blue and gold"

# Multimodal LLMs that generate images (Nano Banana, Gemini 3 Pro Image, GPT-5 image)
ai-gateway image -m google/gemini-2.5-flash-image "a hedgehog wearing a tiny hat"
ai-gateway image -m google/gemini-3-pro-image -o cover.png "minimalist mountain logo"

# Browse models
ai-gateway models --type image
ai-gateway models --search claude
ai-gateway models --json | jq '.[] | select(.type=="language") | .id'

# Video (multi-minute job, shows a spinner)
ai-gateway video "a wave crashing on rocks at sunset" --duration 5
ai-gateway video -m google/veo-3.1-fast-generate-001 --aspect 9:16 --resolution 1080p \
  -o vertical.mp4 "a hummingbird at a flower, slow motion"
ai-gateway video --json --duration 5 "a tiny ripple in water"

Video generation requires a minimum $10 balance on your AI Gateway account.

Claude Code skill

There's a companion Claude Code skill that lets any agent in your terminal use this CLI to generate text or images. Install it with:

npx skills@latest add Vesely/skills/ai-gateway

Source: https://github.com/Vesely/skills/tree/main/ai-gateway.

Key resolution order

1. --key <value> flag
2. AI_GATEWAY_API_KEY env var
3. ~/.config/@vesely/ai-gateway-cli/config.json
4. Interactive prompt (TTY only)