From personal second brain to team brain: one table, labels, and MCP access
Cole Medin · youtube · 2026-09-17
Cole Medin's new tutorial: a personal AI second brain works for one person, but a team needs a different layer underneath — the "team brain."
Architecture: information from chat platforms, repos, and docs flows into a single table; keyword + semantic search over it; every row is labeled at ingestion; the database enforces those labels on every read for security and permissions. Your coding agent queries it over MCP with its own token, while your personal second brain keeps its memory and personality on top.
Chapters: second brain → team brain, keep your agent share the knowledge, one table for every source, the MCP server, labeling at ingestion, access keys and row-level security (demo: same question, different access), search at scale.
Demoed on Oracle AI Database but the structure works with existing tools; full code is open-sourced on the team-brain branch of the Oracle AI Developer Hub and runs on a free container. Note: Oracle-sponsored content, though the architecture is generic.
More from coding & agent
- TypeSafe AI cookbook: line-by-line semantic search scoring 218 lines in one request — hackgoofer · 2026-09-17
- 'Vibe automating': offloading admin work — lead scoring, reporting, email workflows — to AI — aronkor · 2026-09-17
- Why AI coding assistants keep generating deprecated functions, and how devs cope — sharificles · 2026-09-17
- Bend2 launches tomorrow: Victor Taelin's 'post-AGI' language ditches interaction nets for speed — Birchlabs · 2026-09-17
- Devin's Fable 5.1 + SWE-2 Fusion Mode Runs 3+ Hours on Just 1% of Usage, Early User Reports — CtrlAltDwayne · 2026-09-17
- Keeping vLLM's prefix cache warm between agent turns: an engineering guide — bolts98 · 2026-09-17