Scans URLs for phishing, malware, and other threats before your agent clicks through, with optional intent matching to verify the link actually leads where your task expects. Exposes `url_scanner_scan` and `url_scanner_scan_with_intent` tools over streamable HTTP. Works in stateless mode by default at preclick.ai/mcp with a 100 request/day trial tier, no key required. Task-augmented mode queues scans asynchronously and returns a task ID you poll for results. Also ships standalone Python and Node clients if you want scan results without managing MCP protocol details. Useful when you're building agents that navigate the web autonomously and need a security layer before following links scraped from emails, search results, or user input.
claude mcp add --transport stdio cybrlab-ai-preclick-mcp uvx preclick-mcp