Diclaimer: I might change my opinion later.
Figma's release of MCP server support is making rounds in the internet. The feature is still in beta, but I wanted to be quick to write down my initial thoughts so I can see later how right or wrong I got it.
Short answer: it's not that clear.
The claim is "more efficient and accurate design-to-code workflow". But what is the actual problem? Is it uneducated or unskilled developer? Inconsistencies in implementation? Slow design-to-development process? All of these issues are very well-known and recognised problems with various different solutions ranging from tools to processes to organisational setups. Will Figma's MCP make these issues go away?
Efficiency aspect is true on surface level when considering the actual integration. If a developer can stay on one application (code editor) to fetch content from other application it can be argued that the developer workflow improves x amount. It's not absolute truth in all cases but it's generally reasonable statement.
On deeper analysis the efficiency argument is not as strong. In case of MCPs, what is often ignored, is the fact that the underlying interpretation layer of the content is still "just" LLM. Although the integration between systems is now easier the existing issues with LLM output accuracy and content validity are not magically fixed. The validation still requires human-in-the-loop.
Regardless of the fact that the technology is still in its early stages some fundamental truths can be stated.
The following statements are true in all cases, in short-term and long-term:
In my previous post From design to dev I outlined "Types of alignment approaches" from design to development. In this "model" Figma MCP falls into category "enforced" which means the alignment is done via fixed libraries, close coupling with tooling and design references. The challenges in this approach includes extensive tooling and reliance on non-standard non-transferrable solutions and skills. Figma MCP in it's current form is peak "enforced" approach. It practically outsources interpretation of design intention to tools. The tools couldn't be more involved in the "handover" than this.
I believe the true hidden cost of this approach is in the setup and maintenance. The setup will lead to continuous oiling of the engine, more specifically tuning and fixing issues related to:
All of the above requires great amount of discipline to get right. Figma, or any vector drawing application for that matter, is not designed to support workflows that requires strict formatting of content and explicit versioning of assets. I am not confident the following requirements are achieavable and maintainable for great majority of product teams. It sure sounds fascinating, but remains unreachable for the most of us.
In the end the question is "is it worth it?" I am strong believer in principle minimising the investments in non-end-user asset creation and maximise the investment in understanding how the output actually works to make it better. Figma MCP is interesting, but at this point I'd leave it that.