🔍 search-mcp
Free, unlimited Google web search for AI agents. Uses your real Chrome browser (with cookies/sessions) via Kimi WebBridge. No API keys. No rate limits.
Agent: "search for 'golang concurrency patterns'"
→ Google → 24 results → Auto-fetch all pages → Clean markdown → Agent
⚡ One-Liner Install
Windows (PowerShell)
irm https://raw.githubusercontent.com/menesekinci/search-mcp/main/install.ps1 | iex
macOS / Linux
curl -fsSL https://raw.githubusercontent.com/menesekinci/search-mcp/main/install.sh | bash
Go install
go install github.com/menesekinci/search-mcp@latest
search-mcp setup
🧠 Supported IDEs/CLIs
search-mcp setup auto-creates the MCP config for any of these.
🚀 Levels
{"query": "rust async patterns", "level": "high", "site": "docs.rs"}
🔧 How it works
Agent → MCP stdio → search-mcp.exe → Kimi WebBridge → Chrome → Google
↓
go-readability + html-to-markdown
↓
Clean markdown → Agent
- Pagination: auto-paginates Google until target result count is hit
- Dedup: URL-based within a search, same page never fetched twice
- Always live: no cache, no local database — every call hits Google and the target pages fresh. True to the research spirit.
- Single tab per query: each query drives exactly one Chrome tab — it searches there, then navigates that same tab through each result page in turn. Kimi serves one active tab per session and serializes browser commands, so sequential reuse is as fast as juggling multiple tabs, with less churn.
- Multi-query:
queries: [...]runs each query one after another (with a short randomized delay between them to avoid bot detection), each in its own tab within the shared group.
📦 Tools
🛠️ CLI
search-mcp setup # Interactive setup wizard
search-mcp --version # v0.7.0
search-mcp # MCP server (stdio mode)
📋 Requirements
- Chrome (any recent version)
- Kimi WebBridge extension or Kimi Desktop (for browser automation)
- Windows / macOS / Linux
🔗 Links
- Kimi WebBridge — browser automation daemon
- MCP Specification — Model Context Protocol