Step one: slide a new memory layer in underneath it, quietly, while I keep working
I don't like vendor lock-in, and I don't love how trapped I feel inside the Claude Cowork UI either. It's good software. That's exactly the problem- a year and a half of my business now lives inside one company's product, in a memory system I don't control and can't take with me. I'd rather build the exit before I need it than find out the hard way that there isn't one.
I asked Claude directly whether I could just point its memory at Fulcra instead of its own, so Fulcra became the real thing and Cowork just read from it. The answer was no, for a specific reason: Cowork's memory is wired into the product itself. It loads automatically the second I open a session, on any device, before I've typed a word. Fulcra has no version of that hook. It's a place to store files and records, not a system plugged into Claude's own memory pipeline.
I'll say the honest part plainly, because it matters more than the workaround does: I don't think Claude told me no to protect Anthropic's business. I think it told me no because it was true. But intent doesn't change the outcome. There's no export button on Claude's memory. There's no "point this somewhere else" switch. Whether that's deliberate or just how the product got built, the effect on me is the same- leaving is hard, because there's nowhere for a year and a half of context to go.
If I ever want to leave Claude Cowork, or just stop being fully dependent on it, I have a real problem. Every client relationship, every hiring decision, every deal, every pricing call, all of it lives in a memory I can't export and can't point anywhere else. Walking away without a plan means walking away from all of that too. That's not a business I'm willing to run.
• Stopped trying to redirect Cowork's memory, since that door is closed, and built around it instead.
• Set up a folder in Fulcra that mirrors Cowork's memory, file for file, same structure Cowork already uses internally.
• Had Claude write to that Fulcra folder immediately, every time it writes to its own memory, so the copy stays minutes behind instead of days.
• Added a daily job as a backstop, in case anything gets missed.
• Kept working in Claude Cowork exactly like before. That's the point- I don't have to stop yet.
• Left the actual switch to Hermes for a weekend, when I have real time to test it properly instead of doing it in the cracks of a workday.
Right now, every day I spend working in Claude Cowork is also a day Fulcra quietly fills up with the same context, without me lifting a finger for it. I'm not migrating today. I'm loading the new house while I still live in the old one. When I'm ready- some weekend with nothing else on the calendar- I can point Hermes at Fulcra and start working from there instead. And Claude Cowork's own memory doesn't know any of this is happening. Nothing about it changes as part of this, so I can keep going back to it any time I want, picking up exactly where I left off, for as long as I want.
This is step one of two, and it's the quiet one: get the memory layer in place, let it fill up on its own, keep working like nothing's changed.
Step two is the actual move, whenever I decide to make it, on my own schedule, not because a platform forced my hand or a contract ran out. Either way, I stop needing one company's product to remember my own business for me.
Fulcra was designed by people who get privacy and know the importance of an infrastructure solution that can be the secure private datastore for the rest of your life. Here data is yours, under your control, and only shared with the people and tools you choose to share it with.