[2026.2.2-beta]

Second Public Beta Release

1. Canary Token Detection

Canary Token Injection is a security feature that helps detect when data from one user or session is accidentally exposed to another user. Think of it like a "tripwire" for an early warning system that alerts you to potential data leakage in your AI systems.

• Cross User: Canary from user A appeared in response to user B
• Cross Session: Canary from session A appeared in user's different session B
• Memorization: Provider-specific memorization detection
• Stale Canary: Canary older than 7 days appeared (possible memorization)

See the Canary Detection User Guide for more details.

2. Improved MCP Tool Rate limiter

• Improvements to the per-user/per-tool rate limiter for MCP Server PRoxies
• Improved OAuth MCP Tool permission overrides per-user

3. Per-Proxy System Prompt Injection

• Assign System Prompts to individual proxies per team

4. Bug fixes & UI Improvements

• Improved visual indicators for Proxy audit features (enabled or disabled)
• Moved all MCP/LLM Dashboard API Endpoints behind Auth
• Global AI Security Gateway User-Agent applied to all tool requests