The Offsend browser extension is used to protect sensitive information in browser-based AI chat scenarios. When you enter prompts or attach files in the web interfaces of ChatGPT, Claude, Gemini, Grok, Perplexity, or DeepSeek, the extension checks the content before sending and replaces detected sensitive values with placeholders, thereby preventing keys, tokens, private keys, email addresses, and personal data from being leaked to AI services.
Main capabilities
- Local detection and masking: Identify common sensitive data such as API keys, tokens, private keys, email addresses, phone numbers, credit card numbers, IP addresses, UUIDs, and more locally in the browser, and mask them before sending.
- Pre-send review: Display all sensitive matches; you can uncheck values you want to keep as-is, and only checked values will be replaced with placeholders.
- Action options: You can cancel sending, send anyway, or mask and send.
- Built-in scanner toggles: Detectors for email addresses, phone numbers, API keys, tokens, private keys, credit cards, IP addresses, UUIDs, and more can be toggled separately.
- Reaction modes: Configure how Offsend behaves when sensitive data appears in a prompt, including warn (review before sending), auto-mask (mask and notify), and block (stop before masking).
- Placeholder readability: Sensitive values become placeholders like {{API_KEY_1}}, keeping prompts readable.
- Local restoration: The encrypted mapping is stored locally, the time window is your choice, and the original values can be restored before the timer expires. The official description emphasizes that restoration is only a convenience feature and should not be regarded as a password vault.
- Custom rules: Add JavaScript regular expression patterns that run alongside the built-in detectors to match team-specific formats such as internal IDs, ticket prefixes, or proprietary tokens. When entering patterns, the /.../ delimiters are not required, and they run locally along with the rest of Offsend's detection.
How to use On supported browser AI chat pages, enter prompts or attach files as usual, and the extension will check before sending. When a detector hits sensitive data, it can warn, auto-mask, or block sending according to the mode you set. You can also review matches individually before sending, uncheck values you do not want masked, and then choose to cancel, send anyway, or mask and send. To match custom formats, enable Custom rules under Detectors and enter a regular expression pattern without delimiters.
Use cases Suitable for situations where you need to interact with multiple mainstream AI chat services in the browser while worrying that prompts or attachments may accidentally contain sensitiveindividuals or teams handling sensitive information such as keys and personal data. The browser path is covered by this extension; files, folders, and Safe Paste on macOS are covered by Offsend Desktop; repository policies, ignore rules, and runtime gating in coding agents are managed by the CLI and GitHub Action through .offsend.yml.
Limitations The extension does not replace repository policies. Coding agents will still read workspace files, and .offsend.yml plus CLI hooks are the boundary for that path. If both browser prompts and repository context are a concern, the extension needs to be used together with Check or the CLI. The extension targets mainstream browser AI chat; the specific supported browser range, installation methods, and charging and pricing information are not yet available.