A real conversation with a second brain that four different tools had been writing to. The second question is the one that matters.
A real conversation in Claude
What’s new?
On your side, the installer work is the live thread. The onboarding-ux-hardening branch got a full redesign today, dark editorial rail with the orange glow to match the dashboard, plus the sunglasses icon fix and a re-verify pass on the fix wave. All still on the branch, not merged.
Two things sitting open: PR #330 on second-brain-cloudflare, the Prompt Capsule projection one from an outside contributor, hasn’t moved in the daily review. And today’s X batch is scheduled after you sent posts 1 and 3 back for being too polished to reply to.
How do you know all this? Not all of this was done in Claude.
Your second brain. It’s an MCP server I query, and it isn’t Claude-specific. Anything wired into it writes to the same store.
Every entry carries a source field. The last dozen came back tagged:
codex
claude
claude-desktop
email-icloud
One store.Four tools writing to it. Nothing tied to one assistant.
The installer redesign and the sunglasses icon fix I read as
codex
The re-verify pass was tagged
claude
The PR #330 review came from a session tagged
claude-desktop
I didn’t do any of that work. I just read what those sessions wrote down afterward.
How the source field works
Every tool writes to one store.
The brain is an MCP server. Anything wired into it can save, and nothing is tied to a single assistant.
Each entry records where it came from.
A source field is stored alongside the content, which is why tags like codex and email-icloud survive the trip.
Any tool can read the whole thread.
An assistant queries the same store and can attribute each claim to the session that wrote it, including work it never did.
A real exchange with the maintainer’s own second brain. Private details and third-party names are removed. The four source tags are unchanged.
Set up the connection first. Clients need access to your brain and instructions for when to remember and recall. Second Brain works with saved context; it doesn’t silently read every conversation in every app.