AegisGate 4.0 — Three Products, One Mission: Secure Every AI Interaction

Every day, millions of people type sensitive data into AI tools. A developer pastes a database password into Copilot. A lawyer sends a contract to ChatGPT for summarization. A journalist asks Claude to analyze a document containing source identities.

Most of the time, nothing goes wrong. Some of the time, everything goes wrong. And you never know which time it is until it’s too late.

AegisGate exists to make sure the sensitive data never leaves your device in the first place.

Today we’re announcing three products — all open source, all Apache 2.0, all built on the same detection engine — that cover the full AI interaction surface, from your browser to your IDE to your enterprise infrastructure.

🛡️ AegisGate Lens v0.4.1 — Free Browser Protection for Everyone

Lens is a free browser extension for Chrome and Firefox that detects PII, secrets, XSS, compliance risks, and adversarial prompt injections before you click send on 10 AI chat tools:

• ChatGPT, Claude, Gemini, Microsoft Copilot, DuckDuckGo, Perplexity, Mistral, Grok, DeepSeek, Meta AI

It runs 100% on-device. No account. No telemetry. No data collection. Nothing ever leaves your browser.

How it works: As you type a prompt, Lens checks it locally with 151 regex patterns and a CharCNN-BiLSTM machine learning model (~5ms detection time). If it finds something sensitive — a credit card number, an API key, a prompt injection attack — a banner appears warning you what was detected and giving you three options: Cancel, Edit & Redact, or Send Anyway.

Key stats:

• 100/100 adversarial attack patterns caught by the ML model
• 0% false positive rate on the ML detection layer
• ~5ms detection in Chrome (pure JavaScript, no WASM)
• 504 automated tests
• Extension size: 4.2MB (down from 25MB in v0.2.0)

Install: Chrome Web Store · Firefox Add-ons · Source code

⚡ AegisGate Rampart v0.7.1 — Local Proxy + IDE Plugin for Developers

Rampart protects the AI tools that Lens can’t reach — your IDE, your terminal, your API calls. It runs as a local proxy or as an IDE plugin, intercepting AI traffic from Copilot, Cursor, local LLMs (Ollama, LM Studio), and direct API calls to OpenAI/Anthropic.

Two modes:

1. IDE Plugin — installs in VS Code/Cursor, JetBrains, or any LSP editor. Real-time detection as you type, inline warnings like lint errors.
2. Local Proxy — runs on localhost:8443, intercepts all AI HTTP traffic. MITM block mode stops malicious prompts before they reach the model. CA keys encrypted at rest. Audit logs redacted.

Key stats:

• 1,318 test functions, 80.7% coverage
• 13 release assets (macOS, Linux, Windows, Docker — cosign-signed)
• Same detection engine as Lens: PII, secrets, XSS, compliance, adversarial ML
• Apache 2.0, zero external dependencies for core functionality

Download: GitHub Releases · Source code

🏢 AegisGate Platform v4.1.0 — Enterprise AI Security Gateway

Platform is the server-side gateway for teams and organizations. It sits in front of your AI infrastructure — HTTP APIs, MCP servers, A2A agents, ACP/ANP protocols — and scans every request and response for security risks.

Six pillars of protection:

1. HTTP API Security — scans prompts and responses on any REST API endpoint
2. MCP Protocol — protects Model Context Protocol servers from injection and data exfiltration
3. A2A (Agent-to-Agent) — secures inter-agent communication
4. ACP/ANP — Agent Communication Protocol and Agent Network Protocol protection
5. Response Scanning — scans AI responses for PII leakage, hallucinated secrets, injected content
6. Trust Framework — cryptographic attestation of AI interactions for audit and compliance

Key stats:

• 31 compliance frameworks (HIPAA, GDPR, PCI-DSS, EU AI Act, SOC 2, ISO 27001, NIST CSF, and 24 more)
• 12 SIEM integrations (Splunk, Elastic, Datadog, Sumo Logic, Chronicle, QRadar, Sentinel, ArcSight, Rapid7, LogRhythm, Graylog, Loki)
• 4 SOAR integrations (PagerDuty, Jira, ServiceNow, Custom webhooks)
• Web UI (7 pages, served at /ui/)
• Air-gapped deployment supported
• Single 19.1MB Go binary — no runtime dependencies
• 8,000+ tests, 83.1% coverage
• Performance: 23,578 requests per second sustained, 5,000 concurrent VUs, 0% errors

Deploy: docker run -d -p 8080:8080 ghcr.io/aegisgatesecurity/aegisgate-platform:v4.1.0 · Live Demo · Source code

One Detection Engine, Three Deployment Surfaces

All three products share the same detection corpus — the same PII patterns, the same secret signatures, the same XSS vectors, the same compliance mappings, and the same ML adversarial model. What changes is where the detection runs:

You don’t need all three. If you only use ChatGPT in a browser, install Lens (free, 10 seconds). If you also code with Copilot, add Rampart (free, 2 minutes). If you run AI infrastructure for a team, deploy Platform.

Privacy by Design — Not by Promise

Every AegisGate product follows the same privacy principles:

• Zero data collection — no telemetry, no analytics, no phone-home
• All detection is local — patterns and ML run on your device or your server
• Open source — Apache 2.0, every line of code is auditable
• No account required — for Lens and Rampart, no signup, no email, no tracking
• No “Pro” upsell — Lens and Rampart are free forever, not freemium

We believe security tools should not require trust. Our code is open. Our detection runs locally. We collect nothing — because the best way to protect your data is to never touch it.

Getting Started

All three products are Apache 2.0. No vendor lock-in. No proprietary formats. No hidden behavior.

Built by a solo founder. Proven at scale. Open source forever.

Website · GitHub · Documentation · Pricing