I built SpecMem after noticing how fragmented the agentic coding ecosystem has become. Each coding agent uses its own spec format, context resets frequently and switching tools means rebuilding everything from scratch. SpecMem provides a unified, agent-agnostic memory layer for coding agents. It reads specs from Kiro, GitHub SpecKit, Tessl, Cursor, Claude Code, Codex and others, then builds a semantic memory using LanceDB, ChromaDB etc. It also includes impact analysis, drift detection, selective testing and a web dashboard. GitHub: https://github.com/SuperagenticAI/specmem Docs: https://superagenticai.github.io/specmem pip: pip install specmem Early project and still rough in places. Feedback from HN is welcome.