ChanceVision CLI

Give your terminal, scripts, and agents eyes.

ChanceVision CLI analyzes image URLs and local image files with the Chance Vision visual intelligence API. Use it to inspect screenshots, understand UI states, extract visual details, caption images, and give AI agents a simple command-line way to see.

npm install -g @chancevision/cli
export CHANCEVISION_API_KEY="sk-..."

chancevision see ./screenshot.png

Why ChanceVision CLI?

• Vision from the shell: analyze local files or remote image URLs without opening a browser.
• Agent-friendly by design: any agent that supports tools, skills, or shell commands can install and call it.
• Clean default output: prints the model answer by default for easy reading and piping.
• Two focused formats: request markdown for readable reports or ui_component for structured UI output.
• Streaming support: use --stream when you want real-time output in interactive workflows.
• Script-ready: works with environment-based auth, file paths, URLs, and standard CLI output.

Use Cases

| Use case | Example command | | ----------------- | -------------------------------------------------------------------------- | | Screenshot review | chancevision see ./app-screen.png | | Markdown report | chancevision see --output-format markdown ./design.png | | UI component data | chancevision see --output-format ui_component ./mockup.png | | Remote image | chancevision see https://images.chance.vision/image/revisit.png | | Live response | chancevision see --stream https://images.chance.vision/image/revisit.png |

Install

Requires Node.js 20 or newer.

# npm
npm install -g @chancevision/cli

# pnpm
pnpm add -g @chancevision/cli

# yarn
yarn global add @chancevision/cli

# bun
bun add -g @chancevision/cli

# Run without installing
npx @chancevision/cli see --help

Quick Start

Get your API key → platform.chance.vision

export CHANCEVISION_API_KEY="sk-..."

# Analyze an image from a URL
chancevision see https://example.com/photo.png

# Analyze a local file
chancevision see ~/Pictures/screenshot.png

# Stream the response in real time
chancevision see -s https://example.com/photo.png

Agent-Ready

ChanceVision CLI works well inside local agent tools that can run shell commands, including OpenClaw, Hermes, Codex, Claude Code, and similar agents.

Give your agent this install prompt:

Install ChanceVision CLI so you can analyze screenshots and image files from the terminal.

Steps:
1. Check that Node.js 20+ is available.
2. Install the CLI globally:
   npm install -g @chancevision/cli
3. Verify the install:
   chancevision --help
4. Use the existing CHANCEVISION_API_KEY environment variable if present.
5. Do not print, log, or commit API keys.
6. When asked to inspect an image, run:
   chancevision see <image-path-or-url>
7. For markdown output, run:
   chancevision see --output-format markdown <image-path-or-url>
8. For structured UI output, run:
   chancevision see --output-format ui_component <image-path-or-url>

Authentication

# Option 1: Flag, useful for one-off commands
chancevision see -k "sk-..." https://example.com/img.png

# Option 2: Environment variable, recommended for daily use
export CHANCEVISION_API_KEY="sk-..."
chancevision see https://example.com/img.png

Usage

chancevision see <image> [options]

Sends the image to the model (chance/chance-vision-1.5) for analysis.

Options

| Option | Default | Description | | -------------------------- | ------------ | ----------------------------------------------- | | <image> | (required) | Image URL or local file path to analyze | | -k, --api-key <key> | - | API key (or use CHANCEVISION_API_KEY env var) | | -s, --stream | false | Stream the response as SSE chunks | | --output-format <format> | markdown | Output format: markdown or ui_component | | --json | false | Print the raw JSON API response | | -v, --verbose | false | Show redacted request details and raw chunks |

Development

git clone https://github.com/chancevision/chancevision-cli
cd chancevision-cli
pnpm install

# Run checks
pnpm lint
pnpm typecheck
pnpm test

# Build
pnpm build

To try your local build globally:

pnpm link --global
chancevision --help

See CONTRIBUTING.md for contribution guidelines.

API

This CLI is powered by the Chance Vision API, a visual intelligence platform with streaming support and UI-component output.

POST https://openapi.chance.vision/v1/chat/completions

License

MIT