Your own team of AI coworkers.

Give each bot a computer, a memory, and real work. They use your tools, work in parallel, and come back when something needs you.

Free local app. Hosted computers when you need them.

Deskfolk
RResearchWorking
Compare our top three suppliers. I need cost, lead time, and risk before lunch.
I'll check the latest sources and ask Ops for our pricing history.
Comparison ready. Northstar is the best value; Alpine is the safer fallback.Report attached · updated just now
Message Research…
Asking Ops…
3 sources checked
Mac appiPhoneTelegramRoutinesGroup workYour model

Work stops waiting for the chat to finish.

Research can keep digging while Builder changes the site and Ops watches the morning queue. You decide what each coworker owns. Deskfolk keeps their files, messages, and memory out of each other's way.

R
ResearchWeb and sources
Working

Compare the five vendors and flag the contract traps.

8 sources checked
B
BuilderCode and browser
Review

The onboarding fix is ready. I opened the app and checked the full flow.

Pull request ready
O
OpsRoutines and alerts
Scheduled

Weekdays at 8:30, prepare the overnight brief.

Next run tomorrow

Do it once. Keep the skill.

Turn on Teach, complete the job the way you already do it, and let a bot follow along. Deskfolk turns the demonstration into a reusable skill, so the next run starts with your process instead of a blank prompt.

  • Record the workflow, not a lecture
  • Review what the bot learned
  • Use it in chats or scheduled routines
Bot is watching and learning 00:42
Quarterly report
1. Filter the current quarter
2. Copy the final totals
3. Send the review draft

Ask the team, not the same bot twice.

Put bots in a group or mention the right coworker. They can hand work over, ask for missing context, and answer when they are free. You see the useful result in your conversation, not a transcript of their private chatter.

ROB
Launch crew3 coworkers
Can we ship the update on Friday?
RResearch
The customer notes are clean. I'm asking Ops to check the support window.
Research handed this to Ops
OOps
Friday works. I moved the maintenance routine and left the checklist with Builder.
Mac

Research sent the finished comparison.

Now
iPhone

Open the result, reply, or check the bot's computer.

Synced
Telegram

Send the next instruction without opening the app.

Delivered

Your coworkers answer where you are.

Start on your Mac, continue from iPhone, or message a bot through Telegram. The conversation stays with the same coworker, including context, attachments, and the work it already completed.

Local bots need your Mac running. Hosted bots stay reachable when it is closed.

One coworker. One computer.

No shared desktop to trip over. Every bot gets a dedicated seat, while you choose where that seat lives.

Available

Local X11

Private Linux desktops on your Mac. Watch, take over, and hand control back. Included with the local app.

Mac on · Deskfolk open
Soon

Local Aqua

Native macOS seats for work inside the Mac apps you already use. If your Mac can do it, your coworker can learn it.

Native Mac automation

Bring the model you trust.

Sign in with your existing ChatGPT subscription or choose a model through OpenRouter. Deskfolk keeps the orchestration, tools, routines, and computers consistent. Your selected provider handles the turn. There is no silent fallback.

ChatGPTOAuthUse your existing account
OpenRouterAny modelPick the model for the job

Your work has boundaries.

Local by default means local. Turn on sync or hosting only for the coworkers that need it.

Local stays local

Local transcripts, skills, settings, and computer profiles remain on your Mac unless you enable a cloud feature.

Encrypted in transit and at rest

Paired devices use authenticated workspace access. Hosted configuration and sleep state are encrypted before storage.

Tenant-scoped access

Each account has its own authenticated workspace, devices, connectors, limits, storage, and billing state.

Read the security overview

Good questions.

What stays local, what stays awake, and how the computers work.

Where do I talk to Deskfolk?

Use the Deskfolk app on Mac, continue the same conversations in the iPhone app, or connect Telegram to a bot. Local bots need your Mac running. Hosted bots remain reachable when your Mac is closed.

Do bots share one computer?

No. Every bot has a dedicated computer, so one coworker cannot move another coworker’s windows or interrupt its desktop. You can deliberately share context in a group, copy workspace files or skills, and configure connector access without merging their live computer sessions.

How does Deskfolk handle my data and privacy?

Local transcripts, skills, settings, and computer profiles stay on your Mac unless you enable sync or hosting. Paired-device traffic is authenticated and encrypted in transit. Hosted configuration and saved computer state are encrypted before storage, and every account is routed to its own authenticated workspace.

How is Deskfolk different from other AI assistants?

A Deskfolk bot has an ongoing role, memory, files, routines, and its own computer. Several bots can work in parallel, use real websites and tools, collaborate in groups, and keep hosted work moving around the clock instead of waiting inside one chat window.

How many bots can I run simultaneously?

The fair-use safety limit is 16 concurrent hosted bots per account. You can create more coworkers and wake them as capacity becomes available. If your workload needs more than 16 active computers at once, contact us for a larger deployment.

Does my Mac need to stay on?

Yes for local bots: the Mac must be on and Deskfolk must be running. A hosted bot wakes on an on-demand computer for messages and routines, saves its encrypted state when the work is quiet, and remains reachable without your Mac.

Which models can I use?

Sign in with ChatGPT OAuth or choose a model through OpenRouter. The provider you select handles the turn, including tool use. Deskfolk does not silently switch providers when the selected one fails.

Can Deskfolk learn the way I already do a task?

Yes. Start Teach, complete the workflow once, then review the reusable skill Deskfolk creates from the demonstration. That skill can guide future chat work and recurring routines without making you explain every click from scratch.

Give the next job somewhere to go.

Start with a local coworker on your Mac. Add hosted computers only when the work needs to stay awake.

Deskfolk coworker organizing project files