Open sourceWe're on Product Hunt today 🎉

Turn your AI agents' files into your team's work

A shared folder every AI agent on your team reads and writes. Real files at real paths, with a link and a full history seconds after they're written.

No more “send me that again?”

Set up BearDrive

Start your agent — Claude Code, Codex, Gemini CLI, Cowork, Hermes — in the folder you want synced, and paste this in.

It installs the CLI, signs this machine in, asks which folder to sync, and registers the hooks that keep it fresh. You approve one link in the browser.

Backed byTechstars

shared/playbook/discovery.md
playbook
discovery.mdone-pager.html
data
win-loss.csv
# Discovery
kept current by the whole team's agents
## Questions that work
What breaks today if this slips a quarter?
Who else has to say yes before this closes?
## Objections
- Security — send the SOC 2 one-pager first
- Price — anchor on seats saved, not seat cost
ClaudeImage
HermesImage
LEAVE-BEHIND
Four reasons teams switch.
One shared folderEvery agent reads and writes the same files.
Nothing is lostEvery version kept, attributed, restorable.
Yours to runOpen source, on your own storage.
Share by linkNo account needed on the other end.
ChatGPTImage
ABCD
dealoutcomereasoncompetitor
northwindwonchampion + speed
acmelostno security reviewgbrain
initechwonexec sponsor
globexlostbudget froze
hooliwonagent workflownotion
soylentlostno championgbrain
CodexImage
4 agents · 3 files · all at once
synced just now · 0 ops

One shared folder, multiple agents. Nobody pasted anything.

Nothing to integrate. If it can write a file, it already works:

Claude CodeClaude CoworkCodexHermesGeminiObsidianGoogle OKFgbrain

…and anything else that writes a file.

Stop pasting files at each other.

It's the Dropbox moment for AI agents. Before Dropbox, you emailed yourself the file. Your agent and your teammate's agent still do.

Without BearDrive26 hours

Bale9:14amQ3 report for Acme report.html

Snow1:47pmNumbers are a month stale.

Bale4:02pmFixed the numbers report-v2.html

Snow9:31pmStill opens with our roadmap.

Bale10:58amReordered it report-final.html

a week later — which one did we send?

With BearDrive90 seconds

Bale9:14amQ3 report for Acme beardrive.ai/acme/report

Snow9:15amMy agent fixed it in place.

same file, on both machines — every version kept

Stop losing versions.Stop hunting for files.Stop waiting on someone else to edit it.

WHAT YOU GET

It's a folder. That's the whole trick.

Real files at real paths, in the folder you already work in — a place to see who changed what, hand someone a link, and get any version back. Your agent starts every task already knowing what your team knows — instead of from an empty folder.

🔒 app.beardrive.ai/p-b12fd122/wiki/research
wiki / research
NameWhenReads
q3-findings.md2:14pm
pricing-notes.html11m
interviews2h
competitors.csvyesterday
positioning.md1d
meeting-notes.md2d
brand-voice.html4d
decisions1w
onboarding2w
archive3w

Who changed what

Every change is attributed — the person, the machine, and the agent session when an agent did it. Kept for good.

A link for anyone

Share any file as a public URL — no account needed on the other end. Revoke it whenever.

Works offline

Keep editing on a train. Everything converges when you're back, without a merge conversation.

WHAT YOU GET · SHARING

The link is born with the file.

Your agent hands you a link the moment it writes. Gated to your team by default — or public, in one more sentence.

Snow — Claude3:42pm

Just got off the Northwind call — update the deal page.

ClaudeUpdated wiki/deals/northwind.md🔗 — added the call notes, moved the stage to Negotiation, pushed the close date to Aug 29.

by default

app.beardrive.ai/p-b12fd122/…/northwind.md

Everyone on the project, nobody else.

when you publish it

app.beardrive.ai/s/1aad3327224e408d

Anyone with the link opens the page — no account. Gone when you revoke.

WHAT YOU GET · INSIGHTS

Your agents trust a doc
nobody has updated since March.

Every file, sized by how much your agents read it and colored by how long since anyone touched it. The red cells are your risk list — heavily used, quietly rotting.

No names, ever. Heat is counts and dates, never the person who read a file — agent reads by device, the same ids History already shows. Every project member sees the same view.

WHY NOT JUST…

Only one of these is a folder.

Each of these does a job BearDrive doesn't try to do — and pairs with it. None of them is the shared folder your team's agents work in.

Memory toolsGitDrive & DropboxBearDrive
Where it livesBehind an MCP serverA repoTheir folder — ~/DropboxAny folder you choose, in place
Your agent reads itRecalled facts, over MCPIf it clones the repoA placeholder, until mirroring is onThey're just files
Who reads whatNo ideaRead heat, per file

See the full comparison, row by row →

CAN YOU DEPEND ON IT

final-v2-FINAL.html was your version control. This keeps every version, forever.

Every change that syncs is a version — kept, attributed, and restorable, so an agent overwriting your doc is an annoyance, not a loss. And none of it is a black box: the whole product is open source, on storage you can own.

Nothing gets overwritten.

SnowMon 9:02am
Snow's ClaudeTue 4:41pm
PriyaWed 11:20am
Priya's ClaudeThu 8:15am
Snow's Claudetoday 2:14pm

Five versions of q3-findings.md. View or restore any of them, in one click.

Nothing is a black box.

your team's machineswiki/Snowwiki/Priyawiki/MarcoBearDrive hubaccounts · historyshare linksS3 · GCS · R2or a plain folder

No cloud credentials ever touch a laptop — and self-hosted, the storage is yours.

SETUP

You don't install it. You ask.

No console to learn, no terminal to open. Your agent does the whole setup, then keeps everyone's folder fresh on its own.

or read what it does first →

Claude Code
You
Follow beardrive.ai/setup to set up BearDrive. Ask me which folder to sync.
Claude
Installing the CLI and signing this machine in — I'll hand you one link to approve. I'd create shared/ here and sync just that. Or a folder you already have?
You
shared/ is good.
Claude
It's empty — want a structure to start from? Docs + decision records, an LLM wiki, or PARA.
You
Docs.
Claude
Done. shared/ is syncing withdocs/ and decisions/, the hooks that keep it current are registered, and here's the project link.

GET BEARDRIVE

Let us run it. Or run it yourself.

Both are free. See what that means.

Managed Beta

free during beta · live now
  • Nothing to pay and no card to enter
  • Zero setup: your agent handles it
  • Storage, backups, support handled

Self-hosted

free forever · your infrastructure · open source
  • The complete server, teams included
  • One binary: bdrive serve
  • S3 / GCS / R2 / MinIO / a folder

$ brew install runbear-io/tap/beardrive

Stop re-explaining your project
to every agent on the team.

Nothing your agents write should need a Slack search to find again. The PRDs, plans, and notes, in one folder every teammate's agent reads.