AIgis - Secure & Optimize AI Interactions

Secure & optimize your AI interactions. Locally mask secrets and optimize tokens in prompts before sending them to LLMs.

AIgis (pronounced like Aegis, the mythological shield) is a "Client-Side Privacy Firewall" that automatically masks Personally Identifiable Information (PII) and optimizes tokens before your prompts ever leave your device. In an era of "Shadow AI," where sensitive data can easily end up in third-party language models if care isn't taken with the data you enter, AIgis ensures that your privacy is protected while allowing you to fully leverage the capabilities of AI. 🛡️ ZERO-TRUST MASKING: AIgis operates entirely within a secure dual-storage environment right in your browser. It scans your prompts and automatically replaces sensitive data—such as Emails, Phone Numbers, IBANs, IP Addresses, File Paths and Secrets (API keys & tokens from AWS, GitHub, Stripe, Slack, Google and many more)—with unique, traceable placeholders (e.g., [EMAIL_1]). You choose exactly what gets masked. 📖 CUSTOM DICTIONARIES: Have internal project names or trade secrets? Define your own custom "forbidden words" in the dashboard. You can also specify powerful Custom Regex Patterns directly in the UI by wrapping your regex in forward slashes (e.g., /EMP-\d{4}/) — with per-pattern flags (i, m, s, u) and built-in safety validation that rejects broken or dangerous patterns before they reach the engine. AIgis intercepts and masks every match as [CUSTOM_1] before sending. 🔔 RULE SUBSCRIPTIONS FOR TEAMS (NEW): Keep entire teams in sync. Host your organization's masking rules as a single JSON file and let everyone subscribe with one link. AIgis pulls the feed automatically (daily, with offline fallback) — strictly one-way: nothing is ever sent to the feed server, no data leaves your browser. Feeds can even centrally manage extension settings, but only with your explicit confirmation, and every managed setting is clearly badged with 🔒 in the dashboard. Untrusted regex from feeds is neutralized, and all patterns pass a ReDoS-safety validation pipeline. Prefer one-time sharing? You can still export your configuration (including subscription links) as JSON. ⚡SEAMLESS RESTORATION, SMART COPY & PEEK MODE: When the AI generates a response using your placeholders, AIgis automatically injects beautiful interactive badges into the text. Hover to reveal your real data, click a badge to copy securely, or simply highlight and copy paragraphs normally - our Smart Copy feature automatically swaps the placeholders back to your real data in your clipboard! Need to see everything at once? Hold the backtick key (the key left of "1" — "^" on German keyboards) to activate "Peek Mode" and temporarily unmask the entire page. 📦 TOON OPTIMIZATION: AIgis uses Token Optimized Object Notation (TOON) to natively compress heavy JSON and code objects before you send them to LLMs, saving you tokens and preventing context limits. 🗄️ AUTO-PRUNING VAULT (DUAL STORAGE): Your raw data never leaves your device. The secure Local Vault tracks your mapping placeholders in strictly isolated, non-syncing storage and automatically prunes expired entries (default 30 days) to reduce local forensic risk. Meanwhile, your general preferences, Custom Dictionaries and subscription list use Sync Storage to seamlessly share your settings across all your devices. ⚙️ SMART DEVELOPER MODE: Switch between 'Strict' mode to mask everything, or 'Developer' mode to smartly ignore safe programming variables like localhost, documentation keys, git hashes or standard paths like ./node_modules/ to avoid breaking system code inputs. 🌐 NATIVE INTEGRATION: Deep context-aware integration seamlessly hooks into the input fields of major AI platforms. Take control of your data privacy. Your data stays yours. The LLM provider only sees what you want them to see. ⚠️ Disclaimer: AIgis is provided "AS IS", without warranty of any kind. While this extension uses pattern matching and dictionaries to mask sensitive data, no regular expression or heuristic is 100% foolproof. Edge cases, typos, or obfuscated formats may bypass the detection engine. You are ultimately responsible for the data you enter into LLM platforms. The developers of AIgis are not liable for any accidental data leaks, privacy breaches, or damages resulting from the use or failure of this software. Always double-check your prompts if they contain highly sensitive or classified information.