webSearchPrime
Searches the web and returns each result's page title, URL, summary, site name and site icon
LABSCO SUMMARY
Reads error screenshots and suggests fixes; plan subscribers only
Screenshots of stack traces and error dialogs are how bugs actually arrive — pasted into a chat by someone who cannot copy the text. diagnose_error_screenshot reads those and suggests where the problem is and how to fix it, and extract_text_from_screenshot pulls text out of code and terminal screenshots. The catch is the gate: none of this is purchasable on its own, so it comes with the coding plan or not at all.
What it is
Four separate servers Z.ai provides to GLM Coding Plan subscribers, counted together here: web search prime, web reader, a GitHub repository reader, and a vision server that runs on your own computer.
What you get
- webSearchPrime and webReader — search the web, then fetch a page with its title, body, metadata and links
- search_doc, get_repo_structure, read_file — search a GitHub repository's knowledge, walk its tree, read a file in full
- ui_to_artifact, ui_diff_check — turn a UI screenshot into code or a spec, and compare two screenshots for visual drift
- extract_text_from_screenshot, diagnose_error_screenshot — OCR tuned for code and terminal output, and error-dialog diagnosis
- understand_technical_diagram, analyze_data_visualization — structured readings of architecture diagrams and dashboards
- image_analysis and video_analysis — general vision, and MP4/MOV scene analysis up to 8MB locally
Requirements
- A GLM Coding Plan subscription: Lite, Pro, Max or a team plan. The servers are not sold separately, and the key must come from the plan; a team plan has its own key.
- Calls are paid from the plan's credits, the same ones your model use draws on. A web search, page read or repository read costs 1.2 credits; Lite gives you 2,000 every five hours.
- The vision server is charged as GLM-5.3-Flash model use rather than per call.
Setup effort
Paste a URL with a key — add the servers you want to your client
USE CASES
01
Debugging from a screenshot
Read the error dialog or stack trace and get a diagnosis rather than retyping it.
02
Understanding an unfamiliar repository
Search its docs, walk the structure, then read the specific file.
03
Turning a design into code
Convert the UI screenshot to code or a design spec, then check the built version against it.
04
Researching with fresh sources
Search, then fetch the page body and links rather than working from the snippet.
13 tools
Listed in project docs. Your client may show a different list.
webReader
Fetches the page at a given URL and returns its title, body text, metadata and list of links
search_doc
Searches a GitHub repository's documentation for a quick picture of the project, its news, recent issues and pull requests, and its contributors
get_repo_structure
Returns a GitHub repository's directory tree and file list, showing how the project is split into modules and folders
read_file
Reads the full contents of a given file in a GitHub repository
ui_to_artifact
Turns a UI screenshot into code, a prompt, a design spec or a plain-language description
extract_text_from_screenshot
Extracts text from a screenshot with OCR, built for code, terminal output, documents and general text
diagnose_error_screenshot
Reads a screenshot of an error dialog, stack trace or log and suggests where the problem is and how to fix it
understand_technical_diagram
Gives a structured reading of a technical diagram, such as an architecture diagram, flowchart, UML or ER diagram
analyze_data_visualization
Reads dashboards and charts and pulls out trends, anomalies and the main business points
ui_diff_check
Compares two UI screenshots and finds visual differences and gaps between design and build, for UI QA
image_analysis
General image understanding for visual content the specialized tools do not cover
video_analysis
Analyzes video scenes in MP4, MOV, M4V and similar formats (local files up to 8MB), picking out key frames, events and main points
BEFORE YOU INSTALL
What to know first
Tools you get
13 tools named in the project's documentation.
Requires
- an MCP client that can reach a remote endpoint
- An API key from your GLM Coding Plan, created on the platform where you bought the plan. Z.ai: the API keys page in the Z.ai console. BigModel: your Coding Plan overview page (team plans: the team plan's page). Use the setup tabs for that same platform and replace your_api_key (Z.ai) or YOUR_API_KEY (BigModel) with it: the three hosted servers send it in a request header, and the vision server (zai-mcp-server) reads it from Z_AI_API_KEY
- Node.js for the local vision server (zai-mcp-server): 18 or newer with BigModel, 22 or newer with Z.ai
INSTALLATION
Connect your client
Z.ai hosts 3 servers, each with its own URL.
Server setup
web-search-prime
ENDPOINT URL
https://api.z.ai/api/mcp/web_search_prime/mcpRequirements
- An API key from your GLM Coding Plan, created on the platform where you bought the plan. Z.ai: the API keys page in the Z.ai console. BigModel: your Coding Plan overview page (team plans: the team plan's page). Use the setup tabs for that same platform and replace your_api_key (Z.ai) or YOUR_API_KEY (BigModel) with it: the three hosted servers send it in a request header, and the vision server (zai-mcp-server) reads it from Z_AI_API_KEY.
web-reader
ENDPOINT URL
https://api.z.ai/api/mcp/web_reader/mcpRequirements
- An API key from your GLM Coding Plan, created on the platform where you bought the plan. Z.ai: the API keys page in the Z.ai console. BigModel: your Coding Plan overview page (team plans: the team plan's page). Use the setup tabs for that same platform and replace your_api_key (Z.ai) or YOUR_API_KEY (BigModel) with it: the three hosted servers send it in a request header, and the vision server (zai-mcp-server) reads it from Z_AI_API_KEY.
zread
ENDPOINT URL
https://api.z.ai/api/mcp/zread/mcpRequirements
- An API key from your GLM Coding Plan, created on the platform where you bought the plan. Z.ai: the API keys page in the Z.ai console. BigModel: your Coding Plan overview page (team plans: the team plan's page). Use the setup tabs for that same platform and replace your_api_key (Z.ai) or YOUR_API_KEY (BigModel) with it: the three hosted servers send it in a request header, and the vision server (zai-mcp-server) reads it from Z_AI_API_KEY.
MCP Clients
Claude Code — Z.ai
CLAUDE CODE — Z.AI · TERMINAL
claude mcp add -s user -t http web-search-prime https://api.z.ai/api/mcp/web_search_prime/mcp --header "Authorization: Bearer your_api_key"
claude mcp add -s user -t http web-reader https://api.z.ai/api/mcp/web_reader/mcp --header "Authorization: Bearer your_api_key"
claude mcp add -s user -t http zread https://api.z.ai/api/mcp/zread/mcp --header "Authorization: Bearer your_api_key"
claude mcp add -s user zai-mcp-server --env Z_AI_API_KEY=your_api_key Z_AI_MODE=ZAI -- npx -y "@z_ai/mcp-server"Use a key from your Z.ai GLM Coding Plan in place of your_api_key. Web search, web reader and zread run on Z.ai's servers; the vision server runs on your own machine through npx and needs Node.js 22 or newer.
Cline — Z.ai
CLINE — Z.AI · JSON
{
"mcpServers": {
"web-search-prime": {
"type": "streamableHttp",
"url": "https://api.z.ai/api/mcp/web_search_prime/mcp",
"headers": {
"Authorization": "Bearer your_api_key"
}
},
"web-reader": {
"type": "streamableHttp",
"url": "https://api.z.ai/api/mcp/web_reader/mcp",
"headers": {
"Authorization": "Bearer your_api_key"
}
},
"zread": {
"type": "streamableHttp",
"url": "https://api.z.ai/api/mcp/zread/mcp",
"headers": {
"Authorization": "Bearer your_api_key"
}
},
"zai-mcp-server": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@z_ai/mcp-server"
],
"env": {
"Z_AI_API_KEY": "your_api_key",
"Z_AI_MODE": "ZAI"
}
}
}
}Use a key from your Z.ai GLM Coding Plan in place of your_api_key. Web search, web reader and zread run on Z.ai's servers; the vision server runs on your own machine through npx and needs Node.js 22 or newer.
Any MCP client — Z.ai
ANY MCP CLIENT — Z.AI · JSON
{
"mcpServers": {
"web-search-prime": {
"type": "streamable-http",
"url": "https://api.z.ai/api/mcp/web_search_prime/mcp",
"headers": {
"Authorization": "Bearer your_api_key"
}
},
"web-reader": {
"type": "streamable-http",
"url": "https://api.z.ai/api/mcp/web_reader/mcp",
"headers": {
"Authorization": "Bearer your_api_key"
}
},
"zread": {
"type": "streamable-http",
"url": "https://api.z.ai/api/mcp/zread/mcp",
"headers": {
"Authorization": "Bearer your_api_key"
}
},
"zai-mcp-server": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@z_ai/mcp-server"
],
"env": {
"Z_AI_API_KEY": "your_api_key",
"Z_AI_MODE": "ZAI"
}
}
}
}Use a key from your Z.ai GLM Coding Plan in place of your_api_key. Web search, web reader and zread run on Z.ai's servers; the vision server runs on your own machine through npx and needs Node.js 22 or newer.
Clients spell the "type" value differently: Claude Code and VS Code use "http", Cline uses "streamableHttp", and Roo Code uses "streamable-http". Claude Desktop does not accept a URL entry in its config file; remote servers go under Settings → Connectors.
Claude Code — BigModel
CLAUDE CODE — BIGMODEL · TERMINAL
claude mcp add -s user -t http web-search-prime https://open.bigmodel.cn/api/mcp/web_search_prime/mcp --header "Authorization: Bearer YOUR_API_KEY"
claude mcp add -s user -t http web-reader https://open.bigmodel.cn/api/mcp/web_reader/mcp --header "Authorization: Bearer YOUR_API_KEY"
claude mcp add -s user -t http zread https://open.bigmodel.cn/api/mcp/zread/mcp --header "Authorization: Bearer YOUR_API_KEY"
claude mcp add -s user zai-mcp-server --env Z_AI_API_KEY=YOUR_API_KEY -- npx -y "@z_ai/mcp-server"Use a key from your BigModel GLM Coding Plan in place of YOUR_API_KEY. Web search, web reader and zread run on BigModel's servers; the vision server runs on your own machine through npx and needs Node.js 18 or newer. In Claude Code on a GLM Coding Plan, web search, web reader and image analysis are already built in server-side — add these only if you want them with other models.
Cline — BigModel
CLINE — BIGMODEL · JSON
{
"mcpServers": {
"web-search-prime": {
"type": "streamableHttp",
"url": "https://open.bigmodel.cn/api/mcp/web_search_prime/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
},
"web-reader": {
"type": "streamableHttp",
"url": "https://open.bigmodel.cn/api/mcp/web_reader/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
},
"zread": {
"type": "streamableHttp",
"url": "https://open.bigmodel.cn/api/mcp/zread/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
},
"zai-mcp-server": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@z_ai/mcp-server"
],
"env": {
"Z_AI_API_KEY": "YOUR_API_KEY",
"Z_AI_MODE": "ZHIPU"
}
}
}
}Use a key from your BigModel GLM Coding Plan in place of YOUR_API_KEY. Web search, web reader and zread run on BigModel's servers; the vision server runs on your own machine through npx and needs Node.js 18 or newer.
Any MCP client — BigModel
ANY MCP CLIENT — BIGMODEL · JSON
{
"mcpServers": {
"web-search-prime": {
"type": "streamable-http",
"url": "https://open.bigmodel.cn/api/mcp/web_search_prime/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
},
"web-reader": {
"type": "streamable-http",
"url": "https://open.bigmodel.cn/api/mcp/web_reader/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
},
"zread": {
"type": "streamable-http",
"url": "https://open.bigmodel.cn/api/mcp/zread/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
},
"zai-mcp-server": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@z_ai/mcp-server"
],
"env": {
"Z_AI_API_KEY": "YOUR_API_KEY",
"Z_AI_MODE": "ZHIPU"
}
}
}
}Use a key from your BigModel GLM Coding Plan in place of YOUR_API_KEY. Web search, web reader and zread run on BigModel's servers; the vision server runs on your own machine through npx and needs Node.js 18 or newer.
Clients spell the "type" value differently: Claude Code and VS Code use "http", Cline uses "streamableHttp", and Roo Code uses "streamable-http". Claude Desktop does not accept a URL entry in its config file; remote servers go under Settings → Connectors.
Cline (older versions) — Z.ai
For older Cline versions that cannot use the Cline — Z.ai setup:
CLINE (OLDER VERSIONS) — Z.AI · JSON
{
"mcpServers": {
"web-search-prime": {
"type": "sse",
"url": "https://api.z.ai/api/mcp/web_search_prime/sse?Authorization=your_api_key"
}
}
}Put your Z.ai API key in place of your_api_key in the address. Web reader and zread connect the same way: add an entry named web-reader or zread, with web_reader or zread in place of web_search_prime in the address.
