Factory's Droid Lets Developers Click and Edit AI Code Without Switching Windows

The feedback loop with an AI coding agent has always had a hidden tax: you generate output, then jump to a browser, a diff viewer, or a PDF reader to actually see what it made, then context-switch back to tell it what to fix. Factory's latest Desktop App update is designed to eliminate that entirely. Droid's output now lives right next to the session that produced it, and you can point at exactly what needs changing without ever typing a file path or section description.

The three-part loop: Preview, Refine, Ship

The update organizes around a tight workflow. Every artifact Droid produces can now be previewed inline:

• Documents, decks, spreadsheets, and PDFs open in a preview tab directly inside the app
• Live websites load in a built-in browser, side by side with the session that built them
• Code changes appear as a full file-by-file diff, so you know exactly what changed before anything ships

Whatever Droid produces, you can preview it directly where you're already working. That sounds like a small convenience, but for anyone who has spent time copy-pasting outputs between windows, it meaningfully changes the pace of iteration.

Design mode: click the thing you want changed

The more interesting addition is design mode. With design mode, you select an element directly in the preview and comment exactly what you want changed, with no need to explain which section or file you mean. It works both on live browser previews and on generated files like PDFs or presentations.

For websites, the flow is: load the live preview, click a component or section, describe the change right there. No need to explain which page, component, or file you mean. Droid infers the full context from where you clicked.

For code, the interaction mirrors a standard pull request review. You comment directly on the diff, line by line, the same way you'd review a teammate's pull request, and Droid picks up your feedback and keeps going. This is a meaningful UX shift: instead of writing natural-language descriptions of code locations, you annotate the actual lines.

What problem this actually solves

The core friction with agentic coding tools today is the review-and-redirect cycle. As AI agents take on longer-running and more complex work, companies are moving away from purely text interfaces toward tools designed to supervise and coordinate them. Factory's design mode is a direct answer to that: structured, visual feedback that reduces the ambiguity in every correction you send back.

Unlike tools that act more like smart typing assistants, Factory is structured around delegation: you hand off work, and it comes back with a full implementation and a diff to review. This workflow is fundamentally different and more aligned with team-based software development. The new preview and design mode features make that delegation loop faster and less error-prone.

Session organization gets an upgrade too

Beyond the preview and editing features, the sidebar now has expanded filtering, grouping, and organization options. The desktop app is built for running multiple Droids at once; each session lives in the sidebar with its own context, progress, and history, so you can start a Droid on a feature build, kick off another on a migration, and check in on either without losing your place. The new organization tools make managing that parallel workload more tractable as session counts grow.

Where it fits and where it falls short

This update is most useful if you are already in a delegation workflow: assigning Droid a task and then reviewing its output. The design mode and inline diff review are strong for:

• Iterating on generated landing pages or marketing sites
• Reviewing and commenting on AI-generated documents or decks
• Doing a PR-style review pass on Droid's code changes before merging
• Running parallel Droid sessions on different features and switching between them cleanly

It is less useful if your workflow is more conversational or you are using Droid for short, single-turn tasks. Droids do not always produce perfect code; like most AI tools, they can hallucinate logic, miss edge cases, or write inefficient solutions, especially if your prompt or repo is vague. The new review tooling helps catch those issues, but it does not eliminate the need for careful human review before shipping.

Availability and cost

Everything is available in the Factory Desktop App on macOS and Windows across all Factory plans. You can download directly for Apple Silicon, Intel Mac, or Windows from the Factory site. The desktop app is available across all Factory plans, and usage is included in existing subscriptions. Pricing starts at $20/month for the Pro plan, with paid plans up to $200/month. Enterprise pricing is custom and requires contacting sales.

If you are already running Droid sessions and finding yourself constantly switching windows to review output, this update directly addresses that. The design mode in particular is the kind of interaction model that makes AI-assisted iteration feel less like prompt engineering and more like working with a fast collaborator.