Projekt

Feature group two: Editable DOM with inline manipulation. Every element generated by an agent is inspectable, selectable, and editable directly through the live preview panel. You can click on any component—like a button or a hero section—and modify its text, styles, or structure without touching code. The DOM is fully interactive, so changes appear immediately, and you can see how they affect the overall layout, responsiveness, and design. This is useful because it bridges the gap between the prompt-driven output and the final desired result. Instead of asking the agent to make small tweaks through multiple prompts (which can be slow and imprecise), you can simply click and edit. This visual refinement empowers designers and front-end developers to achieve pixel-perfect results quickly. It also reduces the need to switch between a design tool and a code editor, as Projekt handles both roles in one place.

Feature group three: Git integration and deployment readiness. After refining the output visually or through prompts, Projekt provides built-in Git controls for committing and pushing changes. The interface shows a staged list of modified files—components, styles, assets—and you can write commit messages directly. With options to commit all, pull, and push, the workflow is streamlined for continuous integration. This capability is crucial because it turns the experimental AI-driven development into a production-ready process. Users can iterate rapidly, test visual changes, and once satisfied, commit their progress without leaving the workspace. The seamless transition from prompting to deployment ensures that Projekt supports the entire lifecycle of a project, from initial idea to shipped product. Additionally, since Projekt works with any folder on your machine, it fits into existing version control workflows without imposing any proprietary format.

Overall workflow: Projekt's approach is a four-step cycle: connect agents, open a project, prompt and preview, then refine and commit. First, you log in and install your preferred AI agents, bringing your own API keys. Second, you point Projekt at any local folder—no setup, no config, no boilerplate required. Third, you describe what you want in natural language, and as the agent writes code, live previews update in real time. The chat panel shows the agent's thinking and progress, and you can stop or redirect at any point. Fourth, you select any element on the preview, edit inline to tweak styles or content, and when satisfied, stage your changes in the built-in Git panel and commit. This cycle repeats until the project is complete. The entire design emphasizes minimal friction: you never have to switch between terminal, browser, and editor because everything lives in one window.

Concrete use cases: Build a landing page for a design tool—describe the hero section with bold headline, waitlist signup form, and a three-step feature section. Projekt generates the components, and you can refine the layout visually. Another scenario: developing a dashboard for a client. You can ask one agent to create the navigation and sidebar, another to build the data tables, and a third to style components, all while previewing the unified result. For existing projects like a portfolio site or API service, Projekt can be pointed at the folder, and agents will analyze the codebase before making changes—no need to explain context from scratch. Designers can prototype a design system by prompting for reusable UI components (Button, Card, Input) and then visually tweaking them to match brand guidelines. In each case, the outcome is a shipped product that is production-ready, with all changes tracked in Git, ready for deployment.

This tool is built for designers who engineer, indie hackers, full-stack developers, and solo builders who operate as a whole team. It runs on macOS and currently offers early access to an alpha version. Users bring their own API keys for Claude Code, Codex, Gemini, or Opencode, and pay only for the tokens they use through their existing subscriptions—there is no additional lock-in or monthly fee for Projekt itself. The workspace is intentionally minimal, with no feature bloat, no panels you'll never open, and no overwhelming settings. As AI tools and models evolve, Projekt adapts by updating its internal architecture to support the latest best practices, ensuring users never have to restart their workflow. The takeaway: Projekt is the visual workspace that unifies prompting, previewing, and shipping, letting you build faster with AI without sacrificing control or quality.