Skip to main content
机器人会跳舞
New Member
Oct 11, 2026 4:54 PM UTC
Question

Request for plugin review reconsideration: local-font loading differs from use_figma

Hello Figma Community Support,

The Figma Help Center assistant directed me here and stated that Community Support specialists can route this case to the plugin review team. Please help forward my request for clarification and reconsideration.

Plugin: [redacted by moderator to remove personal information]
Plugin ID: [redacted by moderator to remove personal information]
Resource: [redacted by moderator to remove personal information]

The rejection message states:
“We're not able to approve apps that overlap with native capabilities in the Figma MCP server. Consider building skills on top of the official use_figma tool.”

I understand the concern about overlapping functionality. However, this plugin was developed to address a local-font loading difference observed in testing on October 11, 2026:

1. An existing native text node used the font family “HYDiShengYingXiongTi”, style “J”.
2. The official use_figma workflow could read the node, but loading its original font with figma.loadFontAsync failed.
3. The same font loaded successfully through the plugin running in Figma Desktop on the same computer.
4. Through the desktop bridge, the text was edited while preserving its original font, range-level font sizes, colors, and editability.

The intended workflow uses the official Figma MCP first. The companion Codex skill switches to the desktop bridge only after an actual font-loading failure. An independent Codex test successfully followed this fallback and completed the text edit.

These findings describe the tested environment; they are not a claim that every official MCP configuration has this limitation.

The plugin connects to a companion service on the user’s computer at http://localhost:4789. Font files are not transferred. The current implementation also supports user-approved JavaScript execution, and execution results may include document data returned to the connected AI tool. Users review and approve that code in the plugin before it runs.

Could the review team please clarify:

• Is there a supported way for official use_figma to load fonts available in the user’s Figma Desktop environment when the MCP workflow fails to load them?
• If this desktop-font gap is not currently covered, would a companion focused on that gap be eligible for Community publication?
• Would narrowing the plugin to font-specific operations address the overlap concern? What other changes would be required for reconsideration?

I can provide detailed reproduction steps and a recorded comparison if requested.

Thank you for your guidance.

The plugin developer

 

3 replies

Tom Reem
Figmate
Figmate
Oct 11, 2026 7:59 PM UTC

Hi ​@机器人会跳舞 — welcome to Figma Forum, and thanks for these details. 🙏🏻

Plugin reviews are tied to your account, so I can’t share specifics in this public thread. I did check on your request associated with case: 2186825, and it’s currently under review with the Platform Support team.

I’ve passed along everything you shared here so they have the full context as they follow up. They’ll contact you at the email address tied to your Figma account, so keep an eye on your inbox (and your spam folder, just in case).

I also removed some personal information from your original post to protect your privacy. Thanks for understanding.

机器人会跳舞
New Member
Oct 23, 2026 3:28 PM UTC

Hi everyone — following up on my original post above with two new data points that I believe are relevant to the review team's ongoing evaluation (case 2186825).

1. The same rejection is now affecting non-AI use cases.

On October 13, another developer posted a very similar experience in the "Report a Problem" board: "Plugin rejected for 'overlapping with MCP server' — but MCP is not accessible for hardware integrations." Their plugin lets an external hardware controller write to the canvas through a localhost companion plugin — no AI involved — and it was rejected with the same boilerplate reason ("overlap with native capabilities in the Figma MCP server"). That thread also documents a structural gap that applies to my case as well:

  • The remote MCP server's OAuth flow is restricted to approved AI clients (Claude, Cursor, VS Code, etc.); custom applications cannot request the mcp:connect scope.
  • The local Desktop MCP server is read-only — use_figma is not exposed there.
  • The REST API cannot write to the canvas.

In other words, for a fallback that must run inside Figma Desktop with Plugin API access, the companion-plugin architecture is currently the only supported path — there is no official alternative to migrate to.

2. My plugin's overlap with use_figma is narrower than the rejection text implies.

To restate the workflow from my original post: the official MCP / use_figma is always tried first. The desktop bridge is only invoked as a fallback when figma.loadFontAsync actually fails for a font that is installed and usable in the user's Figma Desktop environment (tested example: "HYDiShengYingXiongTi", style "J"). The bridge itself performs no AI inference — it executes a small allowlist of text/font operations (list-fonts, check-font, inspect-text, replace-text, apply-font, create-text) against a user-run localhost service, with token-based first-connect authorization, task deadlines, and no font files transferred.

Questions for the review team / Platform Support:

  1. Is there a supported way for use_figma to load local fonts that fail through the remote MCP path, so the fallback becomes unnecessary?
  1. If the gap is real, would a plugin narrowed to font-only operations (discovery, load verification, and range font application — leaving all general text editing to use_figma) be eligible for Community publication?
  1. Could the team share which specific capability was judged as overlapping, so I can narrow the submission precisely rather than guessing?

Happy to provide reproduction steps, a recorded side-by-side comparison, or the diagnostic build I mentioned earlier. I'm also linking the hardware-integration thread since both cases suggest the current "MCP overlap" clause is being applied to any localhost companion architecture, regardless of whether AI is involved — a clarification in the review guidelines would help many developers here.

Thanks again for your time.

Tom Reem
Figmate
Figmate
Oct 25, 2026 8:17 PM UTC

Hello again ​@机器人会跳舞 — thanks for following up with these additional findings. I’ll pass them along to the Platform Support team, along with your feedback on the clarity of the review guidelines.

I can’t share plugin-specific review details here, but your request is now being tracked under case: 2198263, and the team will follow up with you there.

There may be an additional delay over the weekend, but they’ll be in touch with any available information as soon as possible.

In the meantime, this resource: Plugin and widget review guidelines → Review guidelines, has the latest general guidance. The “Business” section covers how the review team approaches situations where there’s overlap. The guidelines are updated over time, so it’s worth checking back.

Appreciate your patience as we follow up.