Overview
Code Prism - Zero Code Exposure for AI Coding Assistants
Code Prism
Zero Code Exposure for AI Coding Assistants
Code Prism is a VS Code extension that provides end-to-end encryption for AI coding assistants using Fully Homomorphic Encryption (FHE). Your code never leaves your machine unencrypted - the AI processes encrypted data without ever seeing your plaintext.
What is Code Prism?
Code Prism is a security-first VS Code extension that enables you to use AI coding assistants like GitHub Copilot with complete code privacy. Using advanced Fully Homomorphic Encryption (FHE), Code Prism ensures that:
- ✅ Your code is always encrypted before it leaves your device
- ✅ AI servers process encrypted data without ever seeing plaintext
- ✅ Only your device can decrypt the AI's responses
- ✅ No code exposure risk from compromised servers or networks
Key Features
🔒 Always-On Encryption
- Automatic end-to-end encryption using FHE
- No toggle - encryption is always active
- Zero configuration required after initial setup
🛡️ Visual Protection Indicators
- Status Bar: Real-time encryption status (
🛡️ Code Prism: Protected) - Chat Badge: Every AI response shows
🔒 Protected by Code Prism - Active Monitoring: See when encryption is actively processing requests
🔑 Client-Side Key Management
- Encryption keys never leave your device
- Stored securely in VS Code's secret storage
- Export/import for backup and device sync
🚀 Seamless Integration
- Works with GitHub Copilot Chat in VS Code
- Access powerful Mirror 70B and 7B models
- Full support for streaming responses
- Tool calling support for agentic workflows
📊 Debug & Verification
- Advanced settings to verify encryption is working
- View encrypted payloads and decrypt manually
- Detailed logging for security audits
How It Works
End-to-End Encryption Flow
Technology
- Algorithm: Mirror Fully Homomorphic Encryption (FHE)
- Key Storage: VS Code Secret Storage API
- Key Generation: Per-device unique keypairs
- Forward Secrecy: Each message uses unique encryption parameters
Use Cases
Software Development
- Write code with AI assistance without exposing proprietary algorithms
- Get AI-powered code reviews without sharing source code
- Refactor legacy code while maintaining confidentiality
Enterprise Environments
- Meet compliance requirements (GDPR, HIPAA, SOC 2)
- Protect intellectual property during development
- Enable AI assistance in high-security projects
Security-Sensitive Projects
- Government and defense contractors
- Financial services and healthcare applications
- Zero-trust architecture implementations
Available Models
| Model | Context | Output | Tools | Best For |
|---|---|---|---|---|
| Mirror 70B | 128K tokens | 8K tokens | ✅ Yes | Complex reasoning, large codebases |
| Mirror 7B | 32K tokens | 4K tokens | ✅ Yes | Quick questions, code completion |
Requirements
- VS Code: 1.91.0 or higher
- GitHub Copilot: Active subscription
- Mirror API Key: Get from https://platform.mirrorsecurity.io
- WebAssembly Support: Required for FHE operations
What's Protected
When using Code Prism, the following data is encrypted end-to-end:
✅ All chat messages with AI assistants ✅ Code snippets shared with AI ✅ File contents when using @workspace or @files ✅ AI responses from the server ✅ Function/tool calls and their parameters
What's Stored Locally
- Client encryption keypair (VS Code secure storage)
- Server public keys (VS Code secure storage)
- API key (VS Code secure storage)
What's NEVER Sent Unencrypted
- Your source code
- Chat messages
- File contents
- Project context
Next Steps
Built by Mirror Security Zero Code Exposure. Maximum Protection.