Mnemo

Memory infrastructure for AI agents — for when store the latest value isn't good enough. Event-sourced, append-only records — nothing gets overwritten. Citation-backed retrieval: every result traces to its source. Hybrid semantic + lexical search at ~150ms. Model-agnostic, US-hosted. 85.2% on LongMemEval-S (reproducible). Try it now: drop-in MCP server, npx -y getmnemo-mcp (Claude Desktop, Cursor, Windsurf, VS Code). Hosted API and self-hosting coming soon.