Pasteguard

PasteGuard is a privacy layer that masks sensitive data before it reaches an AI provider. It replaces names, emails, phone numbers, API keys, passwords and other customer or confidential details with placeholders such as [[PERSON_1]] and [[EMAIL_1]], so the model still understands the task without ever seeing the original values. Where supported, the originals are restored in the response so the user reads a complete answer. The tool covers three ways teams use AI. A browser extension, marked as beta and experimental, works with ChatGPT, Claude and Gemini and offers automatic masking as well as a keyboard shortcut to mask selected text. A proxy mode sits in front of an application or SDK so that calls to OpenAI, Anthropic or other providers are masked on the way out. A third mode protects coding agents such as Codex, Claude Code, Cursor and Windsurf, hiding API keys, passwords and connection strings found in a project, for example in an environment file. A live dashboard shows request counts, the share of requests masked, scan time and which kinds of personal data or secrets were detected, with detection shown across several languages. It is meant for teams that cannot paste raw customer data, client files or secrets into chat tools or model APIs, including developers and support staff. PasteGuard is open source and self hosted, and the user starts it locally. The site has no pricing page and lists no paid plans. The open source release is free to run on the user's own infrastructure.