z-ai-tools

name	z-ai-tools
description	Use this skill for Z.AI services including generating images from text prompts, analyzing images with vision AI, extracting text from images/PDFs (OCR), searching the web, reading/parsing web pages, or researching GitHub repositories. Activates for image generation, vision analysis, document OCR, web search, web reading, and code research tasks.
keywords	["image generation","vision","image recognition","OCR","web search","web reader","code research","GitHub","zai","z-cli"]
topics	["image-generation","vision","ocr","web-search","web-reader","code-research","zai"]
license	MIT

Z.AI Tools

Unified CLI and MCP tools for image generation, vision, OCR, web search, web reading, and GitHub code research via @briansunter/z-cli.

Setup

npx -y @briansunter/z-cli --help

Set API key in ~/.dotfiles/local/.env:

export Z_AI_API_KEY="your-api-key"

Tools

generate_image

Generate images from text prompts. Auto-downloads to current directory.

Parameter	Type	Required	Description
prompt	string	Yes	Text description (max 4000 chars)
quality	enum	No	hd (default) or standard
size	string	No	Dimensions (default: 1280x1280)
outputPath	string	No	Save directory
filename	string	No	Custom filename

Sizes: 1280x1280, 1568x1056, 1056x1568, 1472x1088, 1088x1472, 1728x960, 960x1728. Custom: 1024-2048px, divisible by 32.

layout_parsing

Extract text, tables, and structured content from images and PDFs (OCR).

Parameter	Type	Required	Description
file	string	Yes	URL or local file path
model	string	No	Model (default: glm-ocr)

Supports JPG, PNG (max 10MB), PDFs (max 50MB, 100 pages).

vision

Analyze images using Z.AI vision model (glm-4.6v). Supports local files and URLs.

Parameter	Type	Required	Description
image	string	Yes	URL or local file path
prompt	string	Yes	Question or instruction about the image
model	string	No	Vision model (default: glm-4.6v)
thinking	boolean	No	Enable reasoning mode
maxTokens	number	No	Maximum tokens in response

Supports JPG, PNG, GIF, WebP, BMP. Local files are base64-encoded automatically.

web_search

Search the web with results optimized for LLM consumption.

Parameter	Type	Required	Description
query	string	Yes	Search query
count	number	No	Results count (1-50, default: 10)
domainFilter	string	No	Limit to domain
recencyFilter	string	No	oneDay, oneWeek, oneMonth, oneYear, noLimit

web_reader

Read and parse web pages to markdown or plain text.

Parameter	Type	Required	Description
url	string	Yes	URL to read
format	string	No	markdown (default) or text
noCache	boolean	No	Disable caching
retainImages	boolean	No	Keep images (default: true)
withImagesSummary	boolean	No	Include image summary
withLinksSummary	boolean	No	Include links summary

search_doc

Search documentation and code within a GitHub repository.

Parameter	Type	Required	Description
repo	string	Yes	GitHub repo (owner/repo)
query	string	Yes	Search query
language	string	No	en or zh

get_repo_structure

View GitHub repository directory structure.

Parameter	Type	Required	Description
repo	string	Yes	GitHub repo (owner/repo)
path	string	No	Subdirectory path

read_file

Read a file from a GitHub repository.

Parameter	Type	Required	Description
repo	string	Yes	GitHub repo (owner/repo)
file_path	string	Yes	Path to file

CLI Usage

# Image generation
npx -y @briansunter/z-cli image "A sunset over mountains" --quality hd --size 1568x1056

# OCR
npx -y @briansunter/z-cli ocr ./document.pdf
npx -y @briansunter/z-cli ocr https://example.com/image.png

# Vision - analyze images
npx -y @briansunter/z-cli vision ./photo.png "Describe this image"
npx -y @briansunter/z-cli vision https://example.com/image.jpg "What objects are here?" --thinking

# Web search
npx -y @briansunter/z-cli search "Claude Code MCP server" --count 5
npx -y @briansunter/z-cli search "bun runtime" --recency oneWeek

# Web reader
npx -y @briansunter/z-cli read https://bun.com
npx -y @briansunter/z-cli read https://docs.anthropic.com --format text --no-images

# Code research
npx -y @briansunter/z-cli zread search facebook/react "hooks"
npx -y @briansunter/z-cli zread structure vercel/next.js
npx -y @briansunter/z-cli zread read vercel/next.js package.json

MCP Server

Single unified server exposes all 8 tools:

{
  "z-ai": {
    "command": "bunx",
    "args": ["@briansunter/z-cli", "mcp"],
    "env": { "Z_AI_API_KEY": "${Z_AI_API_KEY}" }
  }
}

Subset modes: mcp image, mcp ocr, mcp vision, mcp web, mcp search, mcp reader, mcp zread.