Browse / Second brain / brain-in-a-box
brain-in-a-box
BrainInBox
A personal second brain for Claude Code — and a shared memory layer for mixed AI + human teams.
Open BrainInBox/brain-in-a-box → Checked against the four rules.
ai-agentsclaude-codegbraingstackobsidiansecond-brainsemantic-memory
Built for one agent
ships CLAUDE.md and nothing for other agents.
In its own words
Markdown vault + auto-capture + semantic search. Composes with gstack, openclaw, Hermes. One command. macOS, MIT.
A personal second brain for Claude Code — and a shared memory layer for mixed AI + human teams. One command to install. Markdown vault (browsable in Obsidian) + auto-capture hooks + semantic memory (GBrain). Composes with any agent stack: gstack, openclaw, Hermes, or plain Claude Code — they all read and write the same vault.
Everyone gets the same engine, with their own data — everything stays local on their machine.
What's inside
It all runs inside Claude Code (terminal/IDE). No bot, no server, no shared cloud.
Install (per person)
Requirements: macOS, Claude Code, a free ZeroEntropy key (embeddings — free account, no card).
Then the onboarding (15 min, fills in your profile):
install.sh also installs Obsidian (if you don't have it) so you can browse your brain visually — open it and "Open folder as vault" → ~/Documents/Brain.
Verify your install any time (safe, runs in a throwaway temp dir — never touches your real brain):
Optional: gstack (23 AI specialists)
Garry Tan's gstack ships 23 Claude Code skills (a "virtual engineering team": CEO, eng manager, designer, reviewer, QA, CSO, release engineer…). brain-in-a-box installs it alongside on request — they compose: gstack's specialists work, brain-in-a-box captures and remembers.
The repository's own description, then 28% of its README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.