One on-device brain every AI app can subscribe to.
SecondBrain quietly captures the actual content of what you do — messages, pages, docs, anything on screen — keeps it encrypted on your Mac, and serves it over MCP. One brain, every app.
Free to get started · no credit card
Ambient memory
A companion on your shoulder that remembers everything — so you never re-explain yourself to your AI.
SecondBrain quietly captures what you do on your Mac and keeps it as your own memory — right on your device. It fills the blind spots every AI has, no matter how many integrations you bolt on. Plug it into Claude, ChatGPT or Raycast, and switch whenever you like. Your memory is yours, and it comes with you.

See SecondBrain in 2 minutes.
A quick walkthrough of the menu-bar app — how it quietly captures what you do, keeps it on-device, and hands the context to Claude.
Attention isn't memory.
Most tools log that you were somewhere. SecondBrain remembers what was actually there.
They log attention
app · window title · URL · timestamp. They know where your eyes went. Re-open it, search it, and the content you actually need is gone.
title: "Inbox — 42 unread"
url: —
at: 14:07
It captures content
The real text — what was said, written, shown. OCR'd screens, browser innerText, doc bodies. The thing you'd actually want an AI to recall.
body: "Ship the v2 build by Fri,
Priya needs the staging link…"
thread: "Launch checklist"
Capture → encrypt → serve.
One quiet pipeline runs on your Mac. Everything before MCP stays local.
Three tools. One brain.
Everything a subscriber needs to orient, recall, and act on your context.
Recall anything
Full-text search across everything you've captured. Ask for it in plain words, get the actual content back.
What's on screen now
The newest events with their full body — the live picture of what you're looking at this moment.
The orient step
Recently-active apps, people, and threads. The fast “where am I, what am I in the middle of” lookup.
Every meeting, transcribed on-device.
Your mic and the room — captured, labeled by speaker, and dropped straight into your brain. Searchable forever. Never leaves your Mac.
Your brain never leaves your Mac.
Never captured
These get skipped before SecondBrain reads the screen:
Build on the brain.
Don't build another capture pipeline. Subscribe to the one that's already running.
One local endpoint
Point your app at SecondBrain's MCP server. It runs on the user's machine, against the brain they already captured.
Skip the hard part
OCR, accessibility, browser capture, redaction, full-text indexing — already solved. You just call three tools.
A shared memory bus
Every app that subscribes reads the same brain. The user captures once; the whole ecosystem recalls it.
A memory bus, not another silo.
Memory that actually recalls.
LongMemEval-S — the standard test for long-term memory in AI assistants. Higher is better, and SecondBrain runs it fully on-device.
Measured on our own LongMemEval-S harness, fully on-device. What do these benchmarks mean? →
Give your AI a memory you actually own.
Install once. Capture quietly. Let every app you trust subscribe to one brain.
Free to get started · no credit card