# ScrapeSearch MCP (DuckDuckGo + Mojeek)

MCP server that provides:

- `search(query)` — scrape SERP results (first page only)
- `fetch(url)` — fetch and return clean content (markdown/text) or raw HTML

## Requirements

- Node.js 18+

## Install & Build

```bash
npm install
npm run build
```

## Global install

```bash
npm install -g .
```

Then run the server with:

```bash
scrapesearch-mcp
```

## Run (STDIO)

```bash
node dist/index.js
```

## Codex CLI config

`~/.codex/config.toml`

```toml
[mcp_servers.scrapesearch]
command = "scrapesearch-mcp"
args = []
env = { LOG_LEVEL = "info" }
```

## Claude Desktop config

`claude_desktop_config.json`

```json
{
    "mcpServers": {
        "scrapesearch": {
            "command": "scrapesearch-mcp",
            "args": [],
            "env": { "LOG_LEVEL": "info" }
        }
    }
}
```

## Notes

- DuckDuckGo is preferred, with Mojeek fallback if results are empty.
- No pagination; only the first page is scraped.
- SSRF protections block private/loopback targets.

## Scripts

- `npm run dev` — run server in dev mode (TS)
- `npm run build` — compile TS → `dist/`
- `npm test` — fixture-based tests (offline)