Current Versions

Recent Changes

V2 Memory System

• Plan-based AI model selection — Free and Personal plans use the Small model for enrichment. Team and above use the Medium model for higher quality extraction
• Optimized diff capture — Reduced from 5,000 to 3,000 characters per commit for better performance
• Evidence truncation — Reduced from 3,000 to 2,000 characters for faster processing
• Quote verification — Programmatic verification rejects memory items with hallucinated citations

MCP Server 3.5.0

• 23 tools available for Claude integration
• V2 brain document loading with ETag caching
• Semantic search via embeddings
• Hygiene agent with two-phase workflow
• Git digest with SHA-based range tracking
• Context pack building with semantic search

VS Code Extension 1.3.0

• Deep link setup from dashboard
• Automatic git commit capture with diffs
• Session watcher with pipeline tracking
• Context tree sidebar
• Multi-tool configuration (Claude, Cursor, Copilot, Windsurf)

CLI

• 17 commands covering auth, memory, context management, and operations
• V2 brain export to multiple formats (cursorrules, copilot, markdown)
• Daemon mode for background auto-capture
• Hygiene agent with interactive review
• Evidence collection and ingest

Roadmap

Features under active development:

• Light mode for documentation
• Additional AI model options
• Webhook event expansion
• Enhanced semantic search
• Mobile dashboard