One memory for several AI tools. An ongoing experiment.

01 · Why

I use several AI tools, and I want to be able to move between them without rebuilding the context every time. Sometimes I start work in one and ask another to pressure-test it. Sometimes I just want to pick up where I left off.

02 · What works

I've been working with shared context across Claude, Codex and Grok. In my day-to-day use, that part works well. I can carry a problem or an idea between tools and keep going.

Each tool also has a job. Claude usually plays devil's advocate on my ideas and writing. When one tool finishes some code, I have another one check it.

03 · The harder part

The harder part comes when we finish. Claude and Codex often need a reminder to update the shared context with what changed. If I forget to ask, something we worked out stays in the conversation and doesn't make it into the next one. The system still depends on me remembering to tell it to remember.

Grok has been more proactive about those updates in my use so far. I haven't established why, but the difference has made me pay closer attention to how each tool maintains the context it receives.

04 · What I'm still working through

That's the part I'm still working through. Shared memory helps me move between tools, but it needs to keep up with the work for that continuity to hold. My next challenge is figuring out how to connect Claude Duet to the same memory.