WebMCP Shield : AI Agent Firewall & Control

Choose which WebMCP tools AI agents may use on each website. Block purchases, single tools, whole sites, or everything.

WebMCP Shield : AI Agent Firewall & Control Take control of autonomous AI agents interacting with your browser. WebMCP Shield is a real-time security firewall and permission manager built specifically for the Chrome WebMCP (Web Model Context Protocol) standard. As websites adopt WebMCP to expose in-page tools such as search, cart updates, form submissions, and financial transactions WebMCP Shield acts as a Zero Trust security layer between in-page AI models and your sensitive web sessions. KEY CAPABILITIES: Automated Purchase & Payment Protection - Intelligently identifies and blocks transactional actions (checkout, pay, buy, charge card, order placement). - Inspects tool schemas for sensitive payment fields (card numbers, CVV, IBAN, billing address). - Recognizes semantic context and negation patterns to distinguish safe lookups from financial commitments. Granular Tool & Domain Permissions - Per-Tool Control: Allow, block, or require manual approval for individual tools on any site. - Site-Wide Rules: Block all WebMCP tools on specific websites or wildcards across entire subdomains. - Global Kill Switch: Pause agent actions or enforce an instant global block across the entire web with one toggle. Cross-Origin & iFrame Defense - Detects when tools are registered by cross-origin embedded frames or third-party providers. - Blocks third-party tools from running under your first-party credentials without explicit permission. Comprehensive Activity & Audit Log - Keep track of every tool discovered, invoked, allowed, or blocked. - Audit tool usage counters to see when and how AI agents attempt to interact with your tabs. PRIVACY & ZERO TRUST BY DESIGN - 100% Local & Private: Runs entirely inside your browser using Chrome extension sandboxing. Zero cloud processing, zero analytics, zero data collection. - Prototype-Hardened: Built with defensive JavaScript isolation to prevent hostile web scripts or malicious pages from tampering with security hooks. - Zero Impact on Normal Browsing: Does not interfere with human clicks, regular extensions, or typical browsing speeds. WHO IS THIS FOR? - Everyday Users & Shoppers: Prevent AI assistants from making accidental purchases or modifying orders without confirmation. - Developers & AI Engineers: Debug and monitor WebMCP tools, test declarative forms, and inspect `modelContext` payloads in real time. - Security & Privacy Advocates: Bring Zero Trust principles to the agentic web. HOW IT WORKS 1. Install WebMCP Shield and pin it to your toolbar. 2. Browse normally. When an AI agent or website activates WebMCP, the shield badge displays active tools. 3. Click the extension icon to view detected tools, toggle purchase locks, or customize rules per website. Defend your browser sessions against unexpected agentic actions. Install WebMCP Shield today!