Automate your Figma handoffs with an AI agent

How do you find one frame in a file with 400 nodes?

By name, server-side.

The search walks the document tree and returns matching node IDs, names, types and the page each sits on: the difference between “somewhere in Checkout v3” and a node you can act on. Search first, then fetch the nodes you need.

Once a node is identified, a companion renders it at the scale you choose. That is how a design reaches a ticket description, a status update or a review doc without anyone taking a screenshot and cropping it badly.

Comments as a work queue, not a graveyard

Twenty-three unresolved comments is a backlog with no owner.

A companion reads every one with its author and the node it is pinned to, and sorts them into the ones that ask a question, the ones that record a decision, and the ones answered in a meeting three weeks ago and never closed.

It can reply too, anchored to a specific node, so the answer lands on the frame it concerns. Posting is a write, so it waits for your approval.

What happens to the frame after you export it?

It goes where the work is: a ticket in Linear or Jira, a spec page in Confluence, a note in a feedback tool, all reached in the same pass. Carrying the export plus its open comments into those is the operations handoff.

Figma AI, as of this writing, works on the canvas itself: variants, bulk renaming, components from a selection. The perimeter work above is what this surface reads, exports and comments for.

Exports expire after roughly thirty days

Exported image URLs are temporary.

Figma’s signed links expire after roughly thirty days, which is fine for a review this week, and for anything permanent the image has to be downloaded and stored somewhere you control.

Team IDs are the other one. Figma exposes no endpoint that lists the teams you belong to, so listing projects starts with a team ID pasted from a team URL. Thirty seconds, once.

Frequently asked questions