AI design for existing code

Design with AI in the context of your real product.

Iterate visually on an existing codebase without pretending the canvas and runtime are the same thing.

Iterel combines a direct Design canvas, project design-system guidance, tracked engine delegation, immutable versions, and an explicit path for reconciling accepted visual work back into code.

Beyond greenfield generation

A real product has components, constraints, and history.

Many AI UI tools are strongest when they generate a fresh page. Existing products are harder: the design must respect established components, tokens, routes, technical constraints, and work already accepted by the team.

Iterel keeps visual work inside the same project record as coding-agent tasks. People can edit directly in Design, or delegate a scoped change after a tracked task exists, without turning an unreviewed prompt into silent canvas state.

The Design workflow

Use AI without giving up visual control.

Direct visual editing

Work with frames, elements, text, images, layout, components, variables, named styles, pages, comments, and presence.

Existing design-system context

Use a repository-extracted system, a catalog choice, or Iterel Base, then give design-touching tasks the resulting design pack.

Tracked AI delegation

A selection-scoped Design request creates or binds a task before an engine is allowed to mutate canvas state.

Scope and provenance

Generated changes are sanitized, bounded to the selected scope, and attributed through their task, agent or tool, and session.

Versions before and after

Accepted Design work keeps immutable before-and-after versions and a review receipt instead of relying on prompt history alone.

A deliberate code path

Design-to-code work is an explicit reconciliation round, so the team can inspect what the engine actually implements.

From visual intent to code review

Keep every generated change accountable.

  1. 01

    Open the existing project

    Start from a local folder, repository-backed project, or the project’s current Design state.

  2. 02

    Establish the design system

    Extract the repository’s system when available, choose a catalog, or use Iterel Base as a defined fallback.

  3. 03

    Select the visual scope

    Choose the frame or elements the next Design request is allowed to affect.

  4. 04

    Track the delegated change

    Bind a task, run the connected engine under the Design mutation contract, and review the resulting receipt and version.

  5. 05

    Reconcile into code

    Create the implementation round, inspect exact Changes and Preview, and accept only the code outcome you reviewed.

Design and runtime stay honest

Two views of the product, with an explicit bridge.

Design frames are inert

Canvas frames are visual snapshots and editable design objects; they do not hide a browser or carry a Play action.

Preview belongs to a task

Desktop Preview runs the exact task worktree with HMR and stops when that managed task runtime closes or changes.

No automatic-sync claim

Iterel does not promise invisible bidirectional synchronization. The reconciliation round makes code implementation reviewable.

Questions, answered

AI visual design for an existing codebase.

Can Iterel work with an existing React or Next.js product?

Yes. Iterel can start from a local folder or repository-backed project, extract available design-system context, and keep Design work connected to tracked implementation tasks.

Does an AI prompt change the Design canvas immediately?

No untracked prompt may mutate Design. Iterel creates or binds a task first, constrains the selected scope, and records versions and a receipt around the change.

Does Design automatically update the codebase?

No. Accepted visual work moves back to code through an explicit Design-to-code reconciliation round, where the resulting files and Preview can be reviewed.

Can I import a Figma frame?

Yes, when the project has an exact usable Figma binding. Figma Import appears under Design +, creates a tracked task receipt, and imports the selected frame rather than granting broad Figma access.

Can I design locally without uploading the project?

Yes. Machine-local projects include Build and Design, and opening Design does not upload the project. Cloud collaboration and review publication are explicit actions.

Explore the workflow

Design the existing product

Give visual iteration the same rigor as code work.

Build and Design are included in unlimited machine-local projects on Free. Opening Design does not upload a local project.