Working with AI¶

The UI and AI assistant use the same command pipeline. As a result, an assistant can build and edit models using the same operations as a person, and N3XD logs each change with the name of the actor who made it.

In-app chat¶

The chat panel sits at the bottom of the viewport, above the bottom bar on mobile. Responses stream as the assistant works, and the model updates when it applies a command. The assistant uses the curated command catalog through the same dispatch() path as the UI. It can also capture a viewport screenshot, including stress results, to check its work. On mobile, the assistant is the only way to edit because the tree and tools are read-only.

Each command appears as a compact trace line. If the assistant repeats the same command several times, such as creating five parameters, N3XD groups those commands in one row, for example parameter_create ×5. Expand the row to see each command; any failures appear in its header. Screenshots have their own lines.

For a task that needs more than a couple of features, the assistant first writes a build plan. The chat shows it as a row such as Plan · 2/5 done. Open the row to read the steps and see which ones are complete; the plan updates as the assistant works.

The assistant can inspect an assembly as well as build one. Ask whether parts collide, and it runs the same interference check available in the UI. It reports which bodies overlap and by how much. It can also combine bodies using union, cut or intersect.

You can attach reference images to a message by dragging them into the chat panel or selecting the attach button. For example, ask the assistant to "model this bracket as shown". You can attach up to four PNG, JPEG or WebP images, each up to 5 MB. N3XD saves them with the conversation and shows them again when you reopen it.

Ask how to do something in the interface, such as "How do I extrude?" The assistant looks up the answer in this documentation and links to the page it used. It can also highlight the matching control with a pulsing ring, just as the guided tutorial does.

The Guided tutorial uses desktop toolbar and sidebar controls, so it is available on desktop only. If you launch it on a phone, a card directs you to a larger screen and leaves your chat unchanged. Start the tutorial from the Guided tutorial card on your dashboard.

Asking for variants¶

To make several versions of a part, ask for them directly. For example, say "make a Small and a Large variant" or "add a version without the mounting tabs". The assistant creates a named configuration for each version and stores its differences as scoped overrides, such as a changed parameter or a feature that is forced on or off. It does not duplicate the part. You can manage the results in the Configurations panel.

Migrating old projects¶

If this version cannot open a .n3xd file, the error notification offers Try to migrate. N3XD opens a new project and gives the file to the assistant, which reads its feature history and rebuilds the model command by command. The migration is best-effort: it does not reproduce imported STEP, STL or IGES bodies, and edge-based selections (such as fillets or chamfers) may land in a different place on the rebuilt solid. Check the result against the original. The assistant reports anything it could not recreate. Migration requires the AI assistant, so it is available only in cloud/SSO mode.

This option applies only when N3XD cannot read the file. If its project id already belongs to another account, the file failed for a different reason and the notification offers Import as copy instead.

Driving N3XD from outside¶

For automation and external assistants, N3XD provides an MCP server. It exposes the same command catalog through execute_command, along with read tools for the model tree, topology, measurements and more. Connect an MCP client to inspect and modify projects programmatically.

The chat, REST API and MCP server all use the same command pipeline, so their modeling behavior stays consistent.