The author moved coding agent memory to Vilix AI's shared memory over MCP to persist context across machines, using semantic retrieval for better accuracy.
I split my coding sessions about 50/50 between my desktop and my laptop, and every switch meant re-explaining the same goddamn project state to the agent. What we touched last session, what we decided about the auth middleware workaround, which hack I'm pretending is temporary (it's not, it's been there since March). I tried the obvious stuff first. Synced a CONTEXT.md through Dropbox and told the agent to read it at session start. Half the time the agent just... didn't. Or it read a stale copy because Dropbox was mid-sync and confidently built on top of three-day-old plans. Fun times. Then I went down the self-hosting rabbit hole for a weekend. Mem0 on a little Postgres box, got it working on the desktop, felt like a genius. Getting the laptop to talk to it reliably ate another evening, and then I realized I had become a guy who maintains a database so his AI remembers things. That was not the plan. What finally stuck was putting the memory outside both machines entirely. I run both setups through Vilix AI's shared memory over MCP. Both clients connect to the same account, the desktop session saves turns there, the laptop pulls them back with get_context. I'm no longer the sync mechanism, which was honestly the part I hated most about all of this. The retrieval is semantic, so it finds what I meant instead of matching my exact wording from last week. And corrections actually stick, last write wins, so when I say "we're not doing that approach anymore" I only say it once instead of updating two machines' notes files like an idiot. Not perfect though. Each client needs its own separate setup, approving one doesn't carry over to the others, so wiring up machine number two was mildly annoying. And it doesn't reach back into old chats on its own, if I forget to save a session that context is just gone. Anyway. If you're bouncing between machines, stop syncing files and put the memory somewhere both clients can actually read. Worked for me after the self-hosted detour ate a perfectly good weekend.
Nexus Memory is an MCP-native memory server that allows AI agents to share context via a unified protocol, enabling persistent and coordinated memory across different agents without custom integration.
The author built HeurChain, a memory broker that provides agent-specific, persistent memory storage for AI agents, surviving restarts and supporting structured and semantic retrieval.
agentmemory is an open-source persistent memory layer for AI coding agents (Claude Code, Cursor, Gemini CLI, Codex CLI, etc.) that uses knowledge graphs, confidence scoring, and hybrid search to give agents long-term memory across sessions via MCP, hooks, or REST API. Built on the iii engine, it requires no external databases and exposes 51 MCP tools.
The author built mem-port, a local MCP server that gives AI copilots (ChatGPT, Claude Code, Cursor, Windsurf) shared long-term memory via embedded SurrealDB, solving context drift between AI tools.
Discusses the challenge of persistent memory for personal AI agents across sessions, comparing setups like Custom GPTs, Mem, and Open Campus's shared memory approach, and asks for community recommendations on handling memory conflicts.