AIVM Brain gives Claude Code a persistent, governed second brain. A plugin hooks into Claude Code's lifecycle, captures what matters from each session, and recalls it in the next one, in any terminal, on any machine. Access is permission-aware, every read and write is recorded, and connection setup is guided inside your workspace.
How do you connect Claude Code?
Connect Claude Code using the current setup instructions inside your AIVM Brain workspace. Public access currently starts with the waitlist.
Get workspace access. Join the waitlist with your work email and confirm your email. If you already have access, log in to your workspace.
Follow the in-app setup for Claude Code. Use the current connection instructions provided in your workspace for Claude Code. Follow the access scope assigned to your agent.
Check access and retrieval. Ask Claude Code a question about an approved source and check its citation. Also test a restricted source to confirm that the agent cannot retrieve content outside its permissions.
CLAUDE.md is not a memory
Claude Code reads CLAUDE.md at startup, and that file is great for stable instructions: build commands, conventions, the shape of the repo. It is also static, per-repository, and maintained by hand. It does not learn. The decision you made on Tuesday about the migration strategy is not in it unless you stopped and typed it there.
A brain is the moving half of that picture. The plugin watches your sessions and captures the durable facts as they happen: the library you chose and why, the flaky test you diagnosed, the deploy gotcha that cost you an hour. Next session, in that repo or another one, Claude Code recalls it instead of rediscovering it.
Across terminals, projects, and teammates
Session memory tied to one machine solves half the problem. The brain is a service, so the same memory follows you from your laptop to a server to a fresh checkout. Key it once per machine and every Claude Code session syncs to the same place.
It is also shared, on your terms. Your teammate's Claude Code can recall the architecture decision your session captured last sprint, if their role clears it. Each person and each agent gets permission-aware retrieval, so the intern's agent does not read the compensation doc, and there is a tamper-evident record of who recalled what. That is the difference between a shared brain and a shared folder.
What gets captured
Durable facts, not transcripts: decisions and their reasons, project state, preferences, hard-won fixes. You can pin facts, correct them, or shred them from the dashboard, and the governed delete actually deletes.