The author critiques Claude's new Projects beta for limited cross-session memory and describes using Vilix AI's shared memory to integrate tools like Claude, Codex, and Cursor for a seamless workflow.
Saw the new Projects beta announcement and my honest first thought was: great, my projects now have a memory inside Claude. My actual projects still live everywhere else. So last month I'm doing this dumb little side thing, a Telegram bot that pings me when a used camera lens I want drops below a price on eBay. Nothing serious. But I'm planning it in Claude on my laptop in the morning and actually coding it in Codex on my desktop at night, because Codex is just faster at boilerplate for me. Every switch means I'm copy-pasting context like it's 2019. One night I pasted the OLD plan into Codex by mistake and it happily built the wrong thing for an hour before I noticed. I sat there staring at it like... who broke this. Oh. Me. The beta fixes exactly half my problem. Sessions inside Claude remember each other now. But my bot lives across Claude, Codex, and occasionally Cursor when I want a diff view. The memory has to live somewhere none of them own, or I'm just collecting vendor-specific project drawers forever. What I do now (and this took me an embarrassingly long Sunday afternoon to set up, I fumbled the OAuth login twice): everything runs through Vilix AI's shared memory. All three tools connect to the same account over MCP, and each one pulls whatever's relevant when I start a session. I don't paste anything anymore. When the plan changes I just say "we're not doing that anymore" once and it's the new truth everywhere, which honestly fixed 90% of my "which version of the plan is current" anxiety. It's not magic though. The models are lazy about saving, half the time I have to nudge them with "save that" or the good stuff never lands in memory. And I definitely set it up wrong the first time so Codex was reading a project I hadn't touched in weeks and confidently asking me about a bot that no longer existed, which was fun. Anyway the beta's neat if you're all-in on Claude, but I'll believe cross-session memory when my projects follow me to whatever tool I'm using at 1am, not just the one with the new feature. Whoever named both features "Projects" owes us all an apology lol
Claude introduces a new Projects feature that coordinates user sessions, provides aggregated status updates with long-lived memory, and is rolling out in beta.
A developer shares their struggle with losing context when switching between Cursor, Claude Code, and Codex, and how they solved it using MEMMY, a tool that syncs chat histories into a shared memory hub for all agents.
Anthropic relaunches the Projects feature in Claude Code, enabling users to manage multiple AI agents in the cloud with shared memory and parallel task execution.
The tweet discusses the new Claude Projects feature, appreciating its cloud-based sessions and project bundling but noting the lack of local thread support for app development.
A developer created an open-source local memory layer called Memmy to share context between Claude Code and Cursor, eliminating the need to re-explain architecture when switching tools.