UpRock for Claude Code

Scrape, search, crawl, and sweep the web from a real global device network.

UpRock turns any website into clean, LLM-ready markdown — with real-browser rendering, anti-bot handling, and per-country geo placement — searches the web across multiple engines and regions at once, and tests uptime and performance from real devices worldwide. It includes an autonomous research agent: describe the data you need in plain language and it finds, navigates, and extracts it across multiple sites.

Unlike datacenter-proxy scrapers, UpRock runs on a distributed network of real mobile and desktop devices around the globe, so requests look like genuine local users — and you can place a crawl or search in a specific region to see exactly what locals see.

Capabilities

Command	What it does
/uprock:scrape	Extract a single page as clean markdown (JS rendering, anti-bot, geo placement).
/uprock:crawl	Crawl multiple pages of a site and return their content as markdown.
/uprock:search	Search the web across multiple engines and countries, deduplicated.
/uprock:sweep	Test a URL's uptime & performance (Core Web Vitals) from real devices across regions.
/uprock:agent	Describe the data you need — the agent autonomously searches, fetches, and extracts it.
/uprock:setup	Configure your API key and verify everything works.

You can also just ask in plain language — "scrape this URL", "search the web for X", "is my site up worldwide?" — and Claude will use the UpRock tools directly.

Setup

1. Install (from a marketplace that lists this plugin):

   /plugin install uprock@<marketplace>
   

2. Get a free API key. Sign up at https://uprock.ai (free credits included), then create a key at https://uprock.ai/app/api-keys. (The MCP setup page at https://uprock.ai/app/mcp also shows your ready-to-use MCP URL.)
3. When prompted at install/enable, paste your key into UpRock API Key. (Or later: /plugin → uprock → Configure.) It is stored securely in your OS keychain and injected into the MCP connection automatically — never written to a file.
4. Run /uprock:setup to smoke-test the connection.

Self-hosted / staging

Set the MCP Base URL plugin option to point at a different UpRock MCP host (default: https://mcp.uprock.ai). The key is appended to the path automatically (<base>/<api_key>/mcp).

How it works

This plugin bundles UpRock's hosted MCP server (Streamable HTTP). Your API key is injected via Claude Code's ${user_config.api_key} substitution into the server URL — https://mcp.uprock.ai/<api_key>/mcp — so the tools appear natively in Claude Code:

• crawl_fetch — fetch a URL through the device network (returns markdown + HTML as crawl:// resources)
• resource_fetch — read the full content of a crawl:// or sweep:// resource
• web_research — multi-engine, multi-region web search
• sweep — multi-region uptime & performance test

The slash commands above are thin orchestration layers over these tools.

Roadmap

• Semantic video search — UpRock's video search (find the exact moment across indexed video by meaning) is available today via the REST API for early-access partners and will be added here once it ships as an MCP tool. Contact developer@uprock.com for early access.

Privacy & security

• Your API key is stored in the OS keychain (Claude Code sensitive user config), never committed or written to project files.
• Requests are made to your configured UpRock MCP host only.

Links

• Website & signup: https://uprock.ai
• API keys: https://uprock.ai/app/api-keys
• n8n community node: https://github.com/uprockcom/n8n-nodes-uprock
• Support: developer@uprock.com

License

MIT © UpRock