Ten objects went into a small Figma migration fixture, and ten came out in Penpot. That count was a useful first check. It did not tell us whether text stayed editable, a component instance still followed its source, or the layout behaved after resizing.
This note describes one recorded test from November 13, 2026. It uses a deliberately small fixture and reports the differences as well as the successful checks. It is not a general compatibility verdict.
The file and the actual import path
The source contained a 600 × 420 board, two native Inter text objects, a fixed 400 × 80 horizontal auto-layout row with 12 px padding and a 16 px gap, two 120 × 40 rectangles, and a simple 100 × 48 component with a linked instance. There were ten non-root objects and one component definition.
The fixture was created through the Figma plugin API. Its properties and component links were read back before export. Penpot Exporter v0.26.3 then exported all pages to a .penpot archive of 13,699 bytes. The archive was imported through Penpot's cloud interface and opened as a separate copy.
This was an exporter-to-native-archive workflow. It did not test opening a .fig file directly in Penpot.
The surprising layout difference
The horizontal Figma row became a Penpot Flex layout with row-reverse. The child order also reversed.
Reading only the direction field would make this look like an immediate ordering failure. The measured positions told a different story: rectangle A remained at (36, 132), rectangle B at (172, 132), and both stayed 120 × 40. The horizontal gap was still:
172 - (36 + 120) = 16 px The visible order and spacing survived this conversion. The underlying layout representation changed.
A separate behavior check widened the fixed row from 400 to approximately 480 through the Penpot plugin API. The two fixed-size rectangles kept their recorded positions, dimensions and gap; the row was then restored to 400. This covers one fixed row. It says nothing about nested layouts, fill or hug sizing.
Text survived, with different measurements
Both strings remained native text objects. Inter, weight 400 and the two recorded font sizes, 28 and 16, were retained. Their bounds changed:
| Text | Figma | Penpot |
|---|---|---|
| Heading | 280 × 34 | 281 × 34 |
| Body | 312 × 19 | 309 × 20 |
To check editing, the heading was temporarily changed through the Penpot plugin API. Its width changed to 412, after which the original text was restored. That establishes an actual edit in this fixture, rather than just the presence of a text string in an exported structure.
The dimension differences also mean this was not a pixel-identical migration. Other fonts, mixed styles and CJK text were outside the test.
A component link needed a behavior check too
The imported main object and instance pointed to the same component definition. A temporary change to the main rectangle's fill, from #1A66B2 to #2D8C51, propagated to the instance. Restoring the original fill propagated as well.
This tested one same-file rectangle fill. It did not cover variants, overrides, cross-file libraries or design tokens.
After reloading the Penpot file, a further read-back confirmed the ten objects, original strings, restored 400 px row, 16 px gap, original color and component links.
A separate export failed
The Penpot MCP export_shape request for a PNG preview returned HTTP 500 / ResourceRequest timeout. The design was visible in the browser, and editor screenshots were saved, but those screenshots do not establish that server-side PNG export worked.
The native archive import and the preview export are separate operations. The first succeeded in this fixture; the second failed. The failure's cause was not established.
A small migration check worth repeating on your own file
Start with a copy containing a few features you depend on. Record source counts, text, dimensions, positions and component relationships. Export and import through the documented route, then inspect both the structure and visible geometry.
Next, make a reversible edit: change text, resize a layout container or change a component property. Check the result, restore the original state, reload, and check again. Test image export separately from import.
This fixture did not cover complex components, prototypes, collaboration, self-hosting, large production files or other versions. A successful small-file check is a starting point for deciding what to test next.
The experiment was recorded for FOSS Desk, which I maintain. The Figma-to-Penpot guide includes these scoped results and their limits. The separate public software list and read-only release checker covers release metadata; it does not contain this migration fixture or its private evidence archive.
AI assistance was used to organize and draft this article from recorded project evidence. The experiment's results and limitations are identified above.


