ai-sdlc-figma-mcp-server

A small, read-only MCP server that wraps the Figma REST API, so an AI SDLC agent gets real MCP tools instead of hand-constructing authenticated HTTP calls. Built for the ai-sdlc-bootstrap's figma provider's mcp_rest connector, but has no dependency on the bootstrap itself — the only project-local file it reads is a consuming project's ai-sdlc.yml.

No Figma Desktop, no Dev Mode, no http://127.0.0.1 dependency — just a FIGMA_ACCESS_TOKEN and providers.figma.target.file_key, read live from the calling project's ai-sdlc.yml at call time (never baked into .mcp.json).

Read-only: every tool issues a GET, nothing else exists to call.

Tools

• figma_get_file_metadata

• figma_list_top_level_frames

• figma_get_node

• figma_get_comments

• figma_export_node_image_url

Related MCP server: figmind

Usage

Run directly via npx — this is how a generated .mcp.json invokes it:

{
  "mcpServers": {
    "figma": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "github:vjackk/ai-sdlc-figma-mcp-server"],
      "env": {
        "FIGMA_ACCESS_TOKEN": "${FIGMA_ACCESS_TOKEN}"
      }
    }
  }
}

Requires ai-sdlc.yml to exist somewhere at or above the process's working directory, with:

providers:
  figma:
    enabled: true
    target:
      file_key: <figma-file-key>

Versioning

Tagged releases (vX.Y.Z) are what the ai-sdlc-bootstrap's provider template should pin to, once this repository has any — npx github:owner/repo#<tag> resolves a specific version; the bare form above tracks the default branch.