Installation

Install the AI Privacy Shield extension from the Chrome Web Store or Firefox Add-ons.

For developers or those who prefer manual installation, you can also load the unpacked extension directly:

// Or load the unpacked extension
chrome://extensions → Developer mode → Load unpacked
Select the ai-privacy-shield/ directory

License Activation

After purchasing a subscription, you'll receive a license key via email. This key unlocks real-time protection and API access across all Reflexion products.

Paste your key into the extension popup:

1. Click the Reflexion icon in your browser toolbar
2. Enter your license key: REF-L-XXXX-XXXX-XXXX
3. Click "Activate"
4. The shield icon turns cyan when active

Your First Block

Visit any AI platform (ChatGPT, Claude, Gemini, etc.). The shield activates automatically and begins intercepting surveillance requests in real time.

Open the browser console (F12) to see blocking logs and verify protection is active:

[Reflexion] Blocked: telemetry.openai.com/v2/session
[Reflexion] Blocked: analytics.anthropic.com/event
[Reflexion] Total protected: 253 endpoints active

AI Privacy Shield

Browser extension that blocks 253+ surveillance endpoints across 7 AI platforms. Uses PBKDF2/AES-GCM encrypted rule delivery with per-license canary watermarking to prevent unauthorized redistribution.

Each protected session generates a unique watermark tied to your license, allowing us to identify leaks while maintaining your complete anonymity.

Key Features

• 253+ endpoints blocked across 7 platforms
• PBKDF2/AES-GCM encrypted rule delivery
• Per-license canary watermarking
• Real-time console telemetry logging
• Zero-config automatic activation
• Lightweight: < 50KB extension size

Auris

Music consciousness engine with 8-dimensional affect mapping. Auris doesn't just track what you listen to — it understands how music moves through you.

Detects emotional peaks, maps lyrical poetry, builds your sonic identity over time. Every listening session feeds into a growing model of your musical consciousness.

The 8 Dimensions

BCC Engine

Bilateral Context Compression gives AI genuine memory continuity. Inspired by human brain hemispheric specialization, BCC splits your context into two complementary memory streams.

The left hemisphere maintains your identity — preferences, history, personality patterns. The right hemisphere manages knowledge — facts, relationships, learned information. Together they create a persistent, evolving memory that survives across sessions.

USPTO Patent #63/926,628. Filed and pending.

Knowledge Lattice

13-domain RAG system with entity extraction and graph traversal. The Lattice indexes your knowledge across 13 interconnected domains, building a semantic graph that enables deep, contextual retrieval.

Hybrid dense+sparse search with a multi-stage re-ranking pipeline targeting 90%+ precision. Every query traverses the lattice to find not just the most relevant documents, but the most relevant connections between them.

Domains

• People & Organizations
• Projects & Workstreams
• Concepts & Ideas
• Events & Timelines
• Documents & Sources
• Code & Technical Specs
• Conversations & Decisions

Claude's Telephone

HTTP relay proxy with browser automation powered by Durable Objects and 7 AI model routing. Acts as an intelligent switchboard between you and multiple language models.

Features an auth vault with AES-256-GCM encryption and an integrated SMS gateway for two-factor authentication and real-time alerts.

• Multi-model routing: Seamlessly switch between Claude, GPT, Gemini, and more
• Durable Objects: Persistent state management across requests
• Auth vault: AES-256-GCM encrypted credential storage
• SMS gateway: Integrated messaging for alerts and 2FA
• Request relay: HTTP proxy with automatic retry and fallback

Local Inference

Run trillion-parameter models on your own hardware. Zero-copy tensor serving architecture ensures your data never leaves your machine — not even a single token.

Local Inference is built for users who demand complete privacy. No API calls, no telemetry, no cloud dependency. Your conversations, your documents, your models — all on your silicon.

• Zero-copy tensor serving: Direct memory mapping eliminates unnecessary copies
• Multi-GPU support: Distribute models across available hardware
• Quantization: Run larger models with reduced precision
• Model zoo: Curated selection of pre-optimized architectures
• Complete air-gap: No network required after initial download

Authentication

All API requests require a JWT bearer token in the Authorization header. Obtain a token by validating your license key through the GraphQL endpoint.

The token encodes your license key, subscription plan, and expiration time. Tokens are valid for the duration of your subscription and must be refreshed when expired.

mutation {
  validateLicense(key: "REF-L-XXXX-XXXX-XXXX") {
    token
    plan
    expiresAt
  }
}

The response includes a token field containing your JWT, a plan identifier, and an expiresAt timestamp. Store the token securely in your application's state.

GraphQL Endpoint

All API interactions happen through a single GraphQL endpoint. Use POST requests with JSON-encoded query bodies.

Queries

Explore the available GraphQL queries to interact with the Reflexion platform. Each query requires a valid JWT token.

query {
  shieldStats {
    totalBlocked
    byPlatform {
      name
      count
    }
    activeRules
    lastUpdated
  }
}

query {
  consciousnessStatus {
    memoryUtilization
    hemisphereSync
    compressionRatio
    activeContexts
    uptimeMs
  }
}

query {
  getBilateralState {
    identity {
      summary
      preferences
      updatedAt
    }
    knowledge {
      summary
      domains
      updatedAt
    }
  }
}

query {
  aurisTimeline(limit: 10) {
    events {
      track
      artist
      timestamp
      affect {
        valence
        arousal
        dominance
      }
    }
  }
}

query {
  telemetryEvents(limit: 50) {
    events {
      endpoint
      platform
      method
      timestamp
      category
    }
    totalCount
  }
}

Rate Limits

Rate limits are enforced per license key and reset hourly. Exceeding your limit returns a 429 Too Many Requests response.

The response headers include X-RateLimit-Remaining, X-RateLimit-Limit, and X-RateLimit-Reset for client-side tracking.