From Personal to Team: A Hands-on AI Knowledge Base Migration Playbook

OwlZealousideal4779 · reddit · 2026-09-27

A maintainer of an open-source desktop AI agent project shares how he upgraded a personal AI knowledge base into shared team infrastructure:

Problem: docs scattered across GitHub, Figma, Slack and local files, multiple versions, no source of truth, no way to grant agents tiered access.
Solution: Notion hosts the docs, links and permissions; WorkBuddy coordinates by pulling from tools, running scheduled jobs, and suggesting updates.
Setup details: an inventory pass organized everything into five Notion databases (product, engineering, design, operations, onboarding), each page linking back to its source (e.g. a GitHub path) so humans and agents share one index. A scheduled job formats Slack product feedback into Notion and produces a daily summary.
Deliberate human-in-the-loop: when comparing raw feedback with known issues, the agent keeps original wording, links matching docs, flags unverifiable items; stale-page detection only suggests edits that a human approves before publishing, so one bad update can't become truth for every agent.
Real results: the agent reproduced an intermittent Windows crash (pulling Electron architecture notes and old issues into a minimal repro plus a fix doc) and assembled material for an overseas roadshow deck.
Caveats: output quality depends on doc hygiene (outdated GitHub material had to be cleaned first), and free Slack/Figma API tiers hit rate limits quickly.

More from coding & agent