edwin-hao-ai/Awareness-SDK

Awareness-SDK attempts to solve the 'amnesia' problem of AI agents across different IDEs (Cursor, Windsurf, etc.) and CLI tools (Claude Code). While the problem is real, the project's current state (8 stars, 0 forks, 36 days old) indicates it is an early-stage personal experiment rather than a production-grade infrastructure tool. The defensibility is extremely low because it relies on intercepting standard API calls—a pattern that is easily broken by SDK updates or bypassed by native platform features. Crucially, this project faces existential risk from two sides: 1) Anthropic's Model Context Protocol (MCP), which provides a standardized way for agents to access local context and memory, and 2) Frontier labs (OpenAI/Anthropic) building native persistent memory into their ecosystems. Competitors like Mem0 (formerly MemGPT) have significantly more traction, funding, and sophisticated RAG/graph-based memory architectures. The 'zero-code interceptor' approach is clever but fragile, making it unlikely to survive as a standalone layer once IDEs and model providers formalize their own cross-session memory protocols. Displacement is likely within 6 months as MCP adoption grows.