Factory.ai drops native desktop app for multi-agent Droids

Most developers either keep their AI tools trapped in a terminal window or stuck inside a flimsy browser tab. Factory is trying to change that.

Factory just rolled out its native desktop app for macOS and Windows, bringing its autonomous Droid agents into a dedicated, multi-agent workspace across all plans at no extra cost.

Instead of juggling endless command-line prompts, you can now run multiple Droids simultaneously right from the sidebar. You can set one agent loose on building a new feature while another tackles a complex database migration, switching between them without losing context or progress.

Giving AI an actual computer to work on

What makes this app genuinely interesting is that your agents don't suffer from amnesia. Factory gives every Droid a persistent computer, meaning your installed packages, cloned repos, credentials, and background services stay intact when you log off.

Teams can choose between managed Cloud Computers with instant sleep and snapshot restores, or turn their own hardware into an agent home base with a single terminal command. That includes local GPU workstations running models via Ollama or vLLM, which Factory says is already powering fully air-gapped setups inside major financial, healthcare, and government institutions.

Crucially, these Droids don't just spit out code inside an isolated sandbox. They actually use your computer like a coworker would.

A Droid can spin up your staging environment in a browser, click through a customer signup flow to see what crashes, jump into VS Code to run extension tasks, and pull metrics from a local spreadsheet. When explaining its work, the agent dynamically decides whether to render Mermaid diagrams, metrics dashboards, or performance flame charts right in the conversation.

It also hooks into the Model Context Protocol, custom skills, and a companion mobile app for checking diffs and approvals away from your desk.

Why it matters

We think Factory's biggest breakthrough here isn't raw model smarts, but expanding who gets to wield autonomous agents.

According to Factory's enterprise numbers, teams adopt Droids twice as fast when both CLI and desktop interfaces are on the table. Even wilder, people who use both environments complete 4.6 times more sessions than CLI-only diehards. It turns out when you give people a visual interface, they simply get more done.

More importantly, it pulls AI agents out of the engineering silo. Product managers can now spin up an agent to draft detailed specifications, and account executives can use them to parse deal memos without ever touching a bash prompt.

The obvious elephant in the room is computer use security. Handing an autonomous agent the keys to your desktop is naturally nerve-wracking. While Factory requires explicit approvals for risky system operations and encrypts all traffic, widespread enterprise comfort will take time.

If you are already running automated dev workflows, the app is an easy recommendation to try today, especially since settings transfer over automatically. But if your organization has strict compliance rules, setting it up on an isolated machine with local models is definitely the smarter first step.