Searchpin
Self-hosted web search for AI agents with no API keys and four engines in parallel
- Pricing
- Free
- Platforms
- CLI, self-hosted, Linux
- Runs
- Locally
- Version
- 1.0.10
- Listing updated
- today
Similar MCP servers
About Searchpin
Searchpin is a free MCP server for web search and scraping. It runs locally on your machine and is started by your AI client.
Searchpin runs on your own machine and queries Baidu, Sogou, Bing CN and Bing International at the same time, which the README says works inside China without a proxy or VPN. Results are merged and re-ranked by a local embedding model so the agent receives a curated list of titles, URLs and snippets rather than a pile of links, and results unrelated to the query are flagged.
A companion fetch tool reads pages in full. It can parse server-rendered pages from frameworks like Next.js and Nuxt, pull out JSON-LD and microdata, and extract text from PDFs locally, with limits of 25 MiB, 200 pages and 2 million characters. Encrypted or image-only PDFs return an explicit error, and OCR is not included.
The server is polite to engines: Baidu searches are spaced at least 20 seconds apart, a Baidu captcha pauses that engine for 10 minutes, and a Sogou challenge pauses it for 60 then 120 seconds. Paused engines are skipped without waiting, and the response reports which engines ran. Setup is pip install searchpin && searchpin-setup, which downloads an embedding model of about 118 MB once. A Docker image and a Python API are also provided.
Install
Run in a terminal
claude mcp add searchpin -- uvx searchpinAdd to claude_desktop_config.json
{
"mcpServers": {
"searchpin": {
"command": "uvx",
"args": [
"searchpin"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"searchpin": {
"command": "uvx",
"args": [
"searchpin"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"searchpin": {
"type": "stdio",
"command": "uvx",
"args": [
"searchpin"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"searchpin": {
"command": "uvx",
"args": [
"searchpin"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"searchpin": {
"command": "uvx",
"args": [
"searchpin"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
What it does
- Four parallel engines
- Baidu, Sogou, Bing CN and Bing International are queried asynchronously.
- Embedding re-rank
- Merged results are ordered by semantic similarity to the query.
- Page and PDF fetch
- Extracts content from server-rendered pages, structured data and text PDFs.
- Engine backoff
- Captcha and challenge pauses are reported with retry times.
- No keys or quotas
- Runs locally with no account, API key or usage limit.
- Compact output
- Search returns titles, URLs and snippets to save context.
Good for
- Give a local always-on agent web search without a billing account
- Search from inside mainland China without a VPN
- Cross-check a claim against several engines
- Read a PDF or JS-rendered page for an agent
- Embed search in a Python script
Pricing
Free
Free, self-hosted, no API keys
Searchpin FAQ
Do I need an API key?
No. There is no account, key or usage limit; the pipeline runs on your own machine.
What is downloaded on first run?
An embedding model of about 118 MB, fetched once through hf-mirror.com.
Does it read PDFs?
Yes, text PDFs up to 25 MiB, 200 pages and 2 million characters. OCR is not included.
How can I run it?
With pip and searchpin-setup, with Docker via ghcr.io/telly6/searchpin, or through the Python API.
Is Searchpin free?
Yes. Searchpin is free to use.
How do I add Searchpin to Claude Code?
Run this command in your terminal: claude mcp add searchpin -- uvx searchpin
What are the best alternatives to Searchpin?
The closest alternatives to Searchpin are Scrapling MCP Server, Firecrawl MCP Server, Douyin_TikTok_Download_API, SearXNG Search and machine_internet.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.



