Integrating Codex with Figma MCP Server

Integrating Codex with Figma MCP Server

Nov 9, 2025 · 2 min read · 297 Words · -Views -Comments

Claude Code (cc) works well with Figma MCP, but the biggest hurdle is the limited weekly quota. Since I also have a ChatGPT Plus subscription, I decided to see if I could bridge Codex with Figma MCP. Although there isn’t much official documentation on this, it is indeed possible and works quite well. Here is the setup guide.

Requirements

  1. Figma Permissions: You need Dev or full Editor permissions.
  2. ChatGPT Plus Membership: Codex is currently not available for free users; I use the GPT Plus tier.

Codex Configuration

Locate your configuration file at ~/.codex/config.toml and add the following:

experimental_use_rmcp_client = true

[mcp_servers.figma_local]
url="http://127.0.0.1:3845/mcp"

In this setup, I am using a locally hosted MCP server, so make sure you have the Figma MCP server running locally.

Once configured, start Codex. If there’s a connection issue with the MCP server, you’ll see an error; otherwise, you’re good to go.

Note: experimental_use_rmcp_client = true is essential for enabling HTTP stream support, which is required to interface with the Figma MCP server.

Usage

  1. Copy Layer Link: In Figma, right-click the desired layer and copy the link.
    Figma Copy Link
  2. Command Codex: Use natural language in Codex to ask for code generation or modifications based on the linked design.

Personal Impressions

  1. Fidelity: In my experience, Codex sometimes offers slightly higher fidelity when translating Figma designs into code compared to Claude. However, for general logic and complex coding tasks, Claude remains superior. Note that neither is perfect—you will still need to manually polish the generated code, but it definitely saves a significant amount of time.
  2. Cost & Quota: A single Claude Pro subscription’s weekly quota is often insufficient for heavy use. If you find yourself hitting the limit, combining GPT Plus with Claude Pro (totaling around 250 CNY/month) provides a robust, multi-model workflow that keeps you productive.
Authors
Developer, digital product enthusiast, tinkerer, sharer, open source lover