AI coworkers with
their own computers.

Message them. Watch them work. Approve.

Accountant watches client deadlines. Today: a VAT return with three bills missing.

Kabout:blank
+
about:blank
d5s07:00
d5s
Opening your computer
  1. Wake-upClient deadlinesdaily 07:00
  2. MemoryKPN bills: portal onlysaved in July
  3. Skillvat-pack-nlteam playbook
  4. ConnectorsMoneybird2 calls · Read · Allow
  5. Its computerKPN Zakelijk portalno API · 3 PDFs
  6. FileQ3 VAT workbookdevries-btw-q3-2026.ods
  7. ApprovalBook 3 billswaits for Femke
Message it in
Watching

In use across industries.
From investment and law firms to software engineering teams.

Investment firm

Screens litigation cases.

Its agents read the case files, keep them in order and follow up with counsel by email.

Law firm

Works in their case system.

It goes through their case-management software and reads every document, like a colleague on the team.

Software engineering team

Reviews every merge.

Its agents check each change for security issues and report what they find to the engineers.

Everything you'd want in a coworker.

Give each agent its own job, and add another when the work grows.

Examples for

Starts before you ask.

It wakes up on a schedule, notices what needs doing and gets going.

Has its own computer.

A browser, files, spreadsheets and a terminal. It uses your apps the way you do.

Works with your team, and each other.

In Slack, Teams or email. Agents bring in other agents, and your team sees every step.

Asks before it acts.

By default, anything that sends or changes data waits for your OK. Reading is allowed.

Gets better over time.

It remembers how your team works, and turns what works into playbooks.

Connectors first.
Its computer when it counts.

Where your tools have an API, it uses it directly: fast and quiet. No API? It opens its own computer and finishes the job.

TeamThrough connectors, where an API existsOn its own computer, where no API exists
MarketingHubSpotSearch ConsolePage indexing report
Customer supportIntercomShopifybol Partner Platform
EngineeringGitHubLinearReal browser on staging
FinanceMoneybirdKPN Zakelijk portal
OperationsShopifyRTM Freight portal
LegalMicrosoft 365CaseDesk + court letter

These are six examples. See all connectors →

Hand off the work.
Not just the questions.

  • Whole jobs, not prompts.Give it the goal. It plans the steps and finishes the work.
  • Works while you sleep.Long jobs run for hours. The result waits for you in the morning.
  • One for every team.Sales, finance, support and engineering. Add an agent when the work grows.
  • Any model.Pick the best model for each job, and switch when a better one ships.

For enterprises

Rolling out agents across a large organisation?

Our engineers build them with your team, on your systems.

d5s for enterprises

Questions.

What people ask after watching it work.

Does each agent really get its own computer?

Yes. Each agent signs in to its own desktop, with its own browser, logins and files. Open it from the chat to watch, or take control.

Do I need the computer if my tools have APIs?

No. It uses connectors wherever an API exists. The computer is for the rest: portals with no API, PDFs, spreadsheets and tools it has to install first.

Which AI models does it use?

Your choice. Pick a model for each agent, from several providers. You can also bring your own model subscription. Talk to us to set it up.

Can it run in our own cloud?

Yes. Use our EU-hosted service, or run d5s in your own VPC or Kubernetes with Helm charts. Talk to us.

Start with one agent.