Working with AI agents in Motif Design

Last updated: November 13, 2026

The Motif agent works inside your live model. Ask it questions, give it tasks, or have it make changes, right from the agent panel. It reads the model you have open, so you can refer to anything you can see.


Open the agent

  1. Open a model.

  2. Click the sparkle icon in the top-right toolbar. The agent panel opens and briefly shows Reading your model…

  3. Type your request in the What would you like to do? box at the bottom of the panel and send it.

💡 Be specific about what you want to change and where. "Add a 3 ft door centered on the north wall of the conference room" works better than "add a door."


What the agent can do

  • Understand your model. Ask "What's in this model?" for a summary, find and count elements, inspect properties, flag missing data, and see who changed what.

  • Build and edit. Create walls, floors, ceilings, roofs, doors, and windows; place furniture and library assets; create levels; and build custom parametric content.

  • Turn a sketch into a model. Drop in a floor plan image or scanned drawing and ask the agent to draw the walls.

  • Document. Set up floor plans across levels, sections, elevations, RCPs, and sheets; add dimensions and notes; and export drawing sets.

  • Work with linked models. Place another project model as a link and change which version you see.

  • Visualize. Generate concept renderings, edit images, isolate or highlight objects, and set sun and shadows.

  • Research and plan. Review client requirements and project documents, research codes, zoning, site conditions, and products, and run calculations or cost comparisons.

The agent only changes your model when you ask it to, and it works within your project permissions.


Organize your work in threads

A thread is a conversation with the agent. Use separate threads to keep separate lines of work organized.

  • Click + Thread to start a new one.

  • Click the clock icon in the panel header to revisit earlier threads.

⚠ Files you drop into a thread, and anything the agent generates there, stay in that thread.

Thread context limits

Each thread can hold a limited amount of information at once, called its context. Your messages, the agent's replies, the steps it takes, and any documents or images it reads all count toward that limit. Reading documents uses the most. Uploading a document to Project Context doesn't use any context until the agent actually reads it.

Motif warns you when a thread reaches 70% of its limit, and again at 90%. When you see the warning, you can:

  • Keep working. When the thread fills up, Motif condenses the earlier parts of the conversation so you can continue. The agent may lose some detail from early in the thread. This works well when you're close to finishing a task.

  • Start a new thread. Click + Thread, give the agent a short summary of where you left off, and re-attach any files you need or use @ to point it at documents in Project Context.

💡 To keep threads from filling up, use one thread per task, @-mention the exact document you need, and ask about the specific section you need rather than the whole document. "What's the required corridor width in section 4 of the program brief?" uses far less context than "Read the program brief."


Give the agent context

  • Project Context. Upload the client brief, specifications, and standards to Project Context so the agent can use them in any thread. Type @ to point it at a specific document. Otherwise the agent picks a document based on its file name.

  • Skills. Type / to run a saved skill, a repeatable procedure the agent can run again.

  • Attachments. Click the paperclip icon, or drag a file into the chat, to share an image or document for this thread only.

  • Connectors. With the connection enabled, the agent can search and read Slack, Microsoft Teams, Outlook, and OneDrive.

See 📄 Use files and skills with the agentfor full details.


Review and undo changes

Everything the agent does is recorded in model history, just like your own edits. Review its changes in the History panel and undo anything you don't want. You can also ask the agent to undo its last change or restore an earlier version of the model.


Good to know

  • Check the results. Walls traced from images, generated renderings, and research are starting points. Verify dimensions and review anything that affects code, life safety, or engineering.

  • Imported geometry has fewer controls. The agent can inspect Revit and Rhino content but can edit less of it than native Motif content.

  • Renderings don't change the model. Generating an image doesn't place anything in your model.

  • No round-trip to Revit or Rhino. Changes the agent makes in Motif don't write back to your Revit or Rhino files.


Need help? Reach out to us at community.motif.io