Trappr — AI Agent Security

Trappr finds security problems in AI agent workflows before they reach production. Open a workflow in n8n and click the Trappr button that appears on the page. The extension reads the workflow from the tab you are already working in and returns a scored report, so you do not have to export a file or paste JSON into another tool. WHAT IT CHECKS Every scan scores the workflow across four areas and explains each finding: Identity and access — webhook endpoints that accept requests without authentication, and credentials that are shared more broadly than the workflow needs. Data protection — API keys and secrets written directly into nodes instead of being referenced from a credential store, and sensitive fields that are sent to a third-party model without masking. Execution safety — paths where untrusted input can reach a language model or a code node, and code nodes that evaluate input dynamically. Visibility — steps that run with no logging or alerting, so a failure or a compromise would go unnoticed. Each finding names the node it came from and what to change to resolve it. HOW SCANNING WORKS ON EACH PLATFORM One-click scanning directly from the page is available for n8n, on both n8n-hosted and self-hosted instances. On Make, Dify and LangFlow the extension recognises the page and shows you the export step for that editor. You then drop the exported file into the popup and it is scanned the same way. Python agent projects can be scanned by dropping a single .py file or a project .zip into the popup. WHAT HAPPENS TO YOUR WORKFLOW Workflows are sent to the Trappr API over HTTPS to be analysed, and the report is returned to the popup. The extension only reads a page when you ask it to scan, and only on the automation platforms listed above. A Trappr account is not required to run a scan. Signing in links scans to your dashboard so you can keep a history, re-scan after changes, and receive alerts.