rohitg00/agentmemory ? reverse-engineered prompt

Build me a TypeScript tool called agentmemory that gives AI coding agents a shared long term memory, so I don’t have to keep explaining the same project rules, decisions, bugs, and preferences every time I switch between Claude Code, Cursor, Codex, Gemini CLI, OpenCode, or other agents.

It should run as one local memory server that agents can connect to through MCP or a normal HTTP API. It should save useful facts from coding sessions, let agents search and recall the right memories later, track confidence and age so stale memories can be updated, and connect related memories so it feels smarter than just notes in a file.

Please include an easy setup flow, clear config, a small live viewer so I can inspect what it remembers, and examples for hooking it into popular coding agents. Keep it local first with no required external database. Add tests, basic benchmarks for recall and token savings, and practical docs so someone can install it and use it quickly. Look up current MCP docs online if needed.