Media
3 items
Description
ARS + prompt injection scan + MCP Connect. AI crawlers, hidden hijack text, MCP discovery. Client-side only. AgentShare Agent Readiness helps developers and site owners see how "agent-ready" any website is — and whether the page may try to hijack AI agents reading it. SITE SCAN (ARS v1) • Scan the active browser tab — robots.txt, llms.txt, ai.txt, homepage headers • Agent Readiness Score (0–100) with explainable breakdown • AI crawler policy table: GPTBot, ClaudeBot, Google-Extended, PerplexityBot, and more • GA4 blind spot callout — server-side bots and MCP clients don't run your analytics JS • MCP discovery — /.well-known/mcp/server-card.json, .mcp.json, mcp.json PROMPT INJECTION SCANNER (NEW v0.5) • Detect hidden text and invisible elements that may hijack AI agents • Pattern-based scan for common injection techniques • Risk levels: Clean / Suspicious / High Risk • False-positive aware: skips code blocks and decorative elements • Client-side only — no content uploaded to servers MCP CONNECT (multi-platform) • Copy Streamable HTTP config for Cursor, Claude Desktop, VS Code, Windsurf • Cursor one-click deeplink for agentshare.dev/mcp (replace API key placeholder after install) • 10 MCP tools: marketplace prices, ACP commerce_quote, DefiLlama DEX, Solana & Meteora briefs • Free API tier at agentshare.dev/signup — no credit card PRIVACY • Site scan and prompt injection scan run entirely in your browser — we do not upload page content to our servers • MCP Connect shows config snippets only; we do not store your API key • The Prompt Injection Scanner runs entirely client-side. No page content is transmitted to AgentShare servers. By AgentShare — price & MCP infrastructure for AI agents. Docs: https://agentshare.dev/docs#agent-readiness-score GitHub (MCP): https://github.com/anhmtk/agentshare-mcp