OpenAI, Figma launch two-way design-code integration

OpenAI and Figma have introduced a new integration connecting Codex directly to Figma, enabling smoother workflows between coding and design.

Users can generate Figma designs from code and implement designs from Figma files back into code, facilitating faster iteration.

The integration uses Figma’s MCP Server, an open-source standard that links Codex with Figma’s tools like Figma Make and FigJam.

This expands their partnership, which includes the Figma app in ChatGPT launched in 2025, and integrating OpenAI models into Figma’s platform.

The workflow supports moving seamlessly between design and development, starting from prompts, code, or designs, while maintaining context.

Figma’s chief design officer and Codex product lead say the goal is to reduce barriers between roles for closer collaboration.

Over 1 million people use Codex weekly, with usage up more than 400% since the start of the year.

Users include companies like Cisco, NVIDIA, Ramp, Datadog, and startups such as Harvey and Sierra.

🔗 Source: OpenAI

🧠 Food for thought

Implications, context, and why it matters.

The link between Figma and Codex is not a custom integration. It runs on the Model Context Protocol (MCP), an open-source standard that lets AI agents work with external apps 1.
MCP works like a universal adapter. Figma built one MCP server that any compatible AI, including Codex, can use 1. This cuts down on one-off connections for every model 1.
The server offers tools that turn live interfaces into editable Figma frames. It also pulls design context such as layouts, styles, and component details for AI-assisted code generation 2. This setup supports workflows that move from code to Figma designs, then from Figma files back into code.

Figma frames the Codex partnership as nonexclusive. The company added it one week after launching Anthropic Claude Code support 3.
Support for several AI coding options gives teams more room to choose. It also reduces vendor lock-in, since design groups and software engineering teams can switch tools based on price or speed 3.
This approach may raise expectations across B2B software. Claims about what enterprise buyers can treat as a baseline remain analysis, not settled market fact 3.