🛡️ AegisGate Lens v0.4.1canonical facts (source: aegisgate-lens repo)
  • 10 AI providers: ChatGPT, Claude, Gemini, Copilot, DuckDuckGo, Perplexity, Mistral, Grok, DeepSeek, Meta AI
  • 5 detection facets: PII (55), secrets (41), XSS (12), compliance (61), ML adversarial (1 Char CNN-BiLSTM model) — 169 regex patterns + on-device ML
  • 504 automated tests: 492/492 Node + 12/12 ML perf/stress
  • 100% adversarial detection (10/10 prompt injections caught by ML model); 2.31% regex FPR on 6,500 WildChat prompts
  • ~0.3ms regex + ~5ms ML (regex p50 0.076ms; ML ~5ms in Chrome, pure JS, no WASM)
  • 100% on-device, zero network egress by default
  • 12 privacy non-negotiables, Apache 2.0, zero external dependencies, zero WASM binaries
  • Free, forever

AegisGate Lens vs AegisGate Platform

Two products from the same security team, designed for different deployment points. This page explains what each does, when to use which, and when to use them together.

TL;DR

  • AegisGate Lens is a free browser extension (also on Firefox) that protects end users browsing AI tools. Browser-level. No account. Privacy-first.
  • AegisGate Platform is an enterprise gateway that protects servers and teams running AI tools. Server-level. Per-team pricing. Central policy + analytics.

Use Lens alone if you’re an individual developer, security researcher, journalist, or privacy-conscious user. Use Platform alone if you’re running AI in production at a company and need central policy + audit + compliance. Use both if you want every layer protected (Lens for users, Platform for servers).


Feature-by-Feature Comparison

FeatureLens (free browser ext)Platform (enterprise gateway)
PriceFree, foreverFree (Community) / $79/mo (Developer) / custom (Enterprise)
Account required❌ No✅ Yes
Browser-side AI provider protection (PII, secrets, XSS, compliance, adversarial ML in ChatGPT, Claude, Gemini, Copilot, DuckDuckGo, Perplexity, Mistral, Grok, DeepSeek, Meta AI)✅ All 5 facets✅ All 5 facets (same corpus)
5-facet detection (PII / Secrets / XSS / Compliance / ML Adversarial). v0.4.1 adds on-device ML.
Privacy: no prompt text, no URLs, no page content sent
Local processing (no server round-trip for default detection)
Ed25519-signed extension bundle + GitHub Actions provenance
Real-time detection (<1ms latency)
MITRE ATLAS (66 techniques)
OWASP LLM Top-10 (5/10 implemented, 24 patterns total)
Custom rules per user
Team-wide policy (one rule applies to all users)
Team-wide analytics dashboard
Central audit log (who saw what, when)
Server-side AI protection (HTTP API scanning for AI services)✅ (176 patterns)
MCP protocol protection (Model Context Protocol)✅ (8 guardrails)
A2A agent-to-agent security (mTLS, capabilities, rate limiting)
ACP (Agent Communication Protocol) security
Response scanning (catches bad AI outputs before they reach the user)
Trust Framework (continuous cryptographically-signed trust scoring)
EU AI Act Compliance Module (120 controls, Professional+ tier)
OWASP compliance module✅ (Developer+)
GDPR / HIPAA / PCI / SOC 2 compliance modules✅ (Professional+)
SIEM export (Splunk, Datadog, Elastic, etc.)✅ (Professional+)
Custom detection rules per team
Slack / Teams / Jira integration✅ (Enterprise)
Self-hosted optionN/A (already client-side)✅ (Enterprise)
Multi-region deploymentN/A✅ (Enterprise)
SLA / 24/7 support✅ (Professional+)

When to use which

Use Lens alone if:

  • You’re an individual developer using AI assistants
  • You’re a security researcher analyzing AI behavior
  • You’re a journalist or writer concerned about prompt privacy
  • You want a privacy-first safety net while you work
  • You’re evaluating AegisGate for your team (start with Lens)

Use Platform alone if:

  • You’re running AI in production at a company
  • You need to enforce AI safety policy across a team
  • You need to log AI interactions for compliance
  • You need server-side AI protection (your own AI services, not just end-user AI tools)
  • You have an SIEM and need alerts

Use both if:

  • You want defense in depth: Lens at the user browser + Platform at the server
  • You need end-user AI protection AND server-side AI protection
  • You’re a security team that needs both behavioral analytics (Platform) and end-user training tools (Lens)
  • You want a Lens-detected threat to be promoted to a Platform-wide policy rule (roadmap item T2.2.2)

Technical comparison

Architecture

  • Lens: Browser extension (MV3). Content script injected into AI provider pages. Detection runs locally in the browser. Bundle signed with Ed25519, distributed via Chrome Web Store.
  • Platform: Self-hosted Docker container (19.1MB). Sits between your AI services and clients. Detection runs in the gateway. Bundle signed with Ed25519, distributed via GitHub Releases.

Detection corpus

Both products share the same detection corpus (the 5-facet taxonomy: PII, secrets, XSS, compliance, adversarial ML). When a new threat is added to one, it’s added to the other.

Privacy model

  • Lens: Privacy-by-default. No telemetry unless explicitly opted in. 12 non-negotiables (see Lens homepage).
  • Platform: Privacy-by-default for detection. Logging is opt-in (Enterprise tier).

Test coverage

  • Lens: 504 automated tests (492/492 Node + 12/12 ML perf/stress). Zero external dependencies, zero WASM binaries.
  • Platform: 8,000+ tests, 83.1% coverage. Zero external dependencies. All tests in go test (Go 1.26+ stdlib).

Source code


Pricing

AegisGate Lens

  • Free, forever. No “Pro” tier. No credit card. No account.

AegisGate Platform

TierPriceBest for
CommunityFreeSolo developers, getting started
Developer$79/moSmall teams, custom rules
Professional$499/moCompliance (HIPAA, PCI, SOC 2, EU AI Act)
EnterpriseCustomSelf-hosted, SLA, 24/7 support

See full pricing for details.


Get started


Questions? Email security@aegisgatesecurity.io (see SECURITY.md for the disclosure policy).