zai-web

name	zai-web
description	Search the web and fetch page content using z.ai MCP tools. Use when you need web search or content extraction. Tools: webSearchPrime (search with filters), webReader (fetch and extract page content), zread (search documentation repos).
metadata	{"openclaw":{"emoji":"🌐","requires":{"env":["ZAI_API_KEY"],"bins":["npx"]},"mcp":{"server":"zai-web","command":"npx -y @z_ai/mcp-server@latest"}}}

z.ai Web Tools

Search the web and fetch content using z.ai MCP tools.

Tools: webSearchPrime, webReader, zread

---

When to Use

✅ USE this skill when:

• You need to search the web for current information
• You need to fetch and extract content from a URL
• You need to search documentation from popular repos
• Brave Search quota is exhausted (backup option)

---

Tools

webSearchPrime

Web search with domain and recency filters.

Parameters:

• search_query — What to search for
• search_domain_filter — Limit to specific domains (optional)
• search_recency_filter — Time-based filter (optional)
• content_size — Max content to return (optional)
• location — Geographic location for results (optional)

Example:

webSearchPrime(
  search_query="OpenClaw memory systems",
  search_recency_filter="month",
  content_size=5000
)

Gotcha: Even with location=us, results can be China-weighted. Use search_domain_filter for English-only sources.

---

webReader

Fetch and extract readable content from a URL.

Parameters:

• url — The URL to fetch
• timeout — Request timeout (optional)
• no_cache — Bypass cache (optional)
• return_format — Output format (optional)
• retain_images — Keep images in output (optional)

Example:

webReader(
  url="https://docs.openclaw.ai/memory",
  return_format="markdown"
)

Use when: You need full page content, not just search snippets.

---

zread

Search and read documentation from popular repositories.

Tools:

• search_doc(repo_name, query, language?) — Search documentation
• get_repo_structure(repo_name, dir_path?) — Browse repo structure
• read_file(repo_name, file_path) — Read specific file

Supported repos: Popular documentation repos (check z.ai for current list)

Example:

search_doc(
  repo_name="openai",
  query="function calling",
  language="en"
)

---

Fallback Strategy

Primary	Fallback	When to Switch
Brave Search (web_search)	webSearchPrime	Brave quota exhausted
webReader	web_fetch (built-in)	If MCP fails

---

Limitations

• API key required: ZAI_API_KEY must be set
• Search bias: Results may be China-weighted; use domain filters
• Rate limits: Unknown — haven't hit them yet

---

Related

• TOOLS.md — Local notes on MCP configuration
• mcporter — MCP server management
• web_search — Built-in Brave Search (primary)
• web_fetch — Built-in content fetcher (backup)