USE CASE · TEAM WIKI

The pages nobody updatesare the ones everyone reads.

For a team whose knowledge lives in documents nobody maintains.The wiki gets written as a side effect of work — and the hub tells you which pages are carrying weight.

Start free →Read the full guide

No credit card. No wiki migration.

  1. 1Someone worksA question gets answered, a decision gets made, a number gets checked.
  2. 2The answer lands in the wikiBecause writing it there is now the cheapest way to finish the task.
  3. 3The next agent reads it firstIt pulls the team's pages before it answers — then writes back what it learned.
and the next task starts from a better wiki

It improves every time anyone works. That is the only wiki maintenance model that survives contact with a busy team.

WRITING IT

Nobody schedules a documentation day.

Work normally. When something is worth keeping, say where it goes — and the page is written, attributed, and on everyone's disk before the turn ends.

Snow — Claude4:41pm

Write up what we learned in wiki/research/q3-findings.md.

ClaudeWritten. Linked it from wiki/index.md so it's findable.

Priya — Claudenext week

Why is churn up in self-serve?

ClaudePer wiki/research/q3-findings.md — single-seat accounts churn 4× faster. I've added this quarter's numbers to the same page.

The second agent read the page and wrote back to it. That is the compounding part — the wiki is not a copy of the work, it is where the work accumulates.

READING ITS HEALTH

Which pages are load-bearing, and which are lying to you.

The hub records what your agents actually read. Plot that against how long since a page changed and the maintenance queue picks itself.

STALE →READS →
quiet and stalenobody's problemhot and staleyour maintenance queuequiet and freshfinehot and freshworking as intended

Usually three pages, not thirty.

Insights (hub admins and org owners) is the quadrant view. Folder listings show read counts to every member, so the signal isn't locked behind an admin screen.

📄pricing.mdedited 14 months ago212 reads / 30d

A page read two hundred times and untouched for a year is not documentation. It's a rumour with a URL.

TRUSTING IT

Every change carries who, which machine, and when.

Every past version is retained — content is stored by hash and never deleted, so a page from eight months ago is still one click away.

🔒 beardrive.ai/p-b12fd122/history?path=wiki/research/q3-findings.md
nowClaude on Priya's machine2 hours ago
v11Claude on Snow's machineyesterday
v10Marco (upload)3 weeks ago
v1Claude on Snow's machine8 months ago

Ask “what changed in the wiki this week?” and your agent answers from the same record.

SET IT UP

Sync the wiki, not the repository.

1

Scope it to the folder that is the wiki.

Snow — Claude10:12am

Set up BearDrive here, but only sync wiki/.

ClaudeSyncing wiki/ only. Everything else in this repo stays local.

Git directories never sync — per-file last-writer-wins would corrupt a repository. A standalone wiki folder works exactly the same way.

2

Leave a map at the root.

📁 wiki/AGENTS.mdresearch/decisions/runbooks/

One short AGENTS.md saying what goes where. Without it, five agents invent five folder structures in a week.

A wiki that gets better because people are busy.

Start free →see the other use cases
What agents read →Read heat, Insights, and what the hub deliberately doesn't record.Shared agent memory →The AGENTS.md map that keeps five agents filing things the same way.Scoping the folder →Subfolder syncing and .bdriveignore, for the parts that shouldn't travel.