Skip to content
agent-ready-repo

figma

Let your agent work with your Figma files from a normal share URL: read a design’s structure, render a frame to an image, pull comments or variables, and turn a FigJam connector diagram into Mermaid. You give it a Figma URL and a frame name — no file keys, node IDs, or subcommands to learn.

What it changes — and what it doesn’t

Section titled “What it changes — and what it doesn’t”

Figma’s REST API is read + comments only, and this pack stays inside that:

  • It does not edit your design. No creating, moving, or changing nodes — that needs Figma’s Plugin API, not REST.
  • It can post comments, and comments are visible to every collaborator on the file. A comment write always asks for your confirmation first.
  • Every render ends with a receipt naming the source frame, the local file it wrote, the format, any skipped or lower-fidelity nodes, and No Figma changes made — so you always know nothing on Figma’s side moved.

Before you install

Section titled “Before you install”

You’ll need:

  1. A Figma account with at least view access to the file you want to read.
  2. A Figma personal access token (PAT), created at Figma → Settings → Security → Personal access tokens.
  3. Python on your machine — the pack shells out to a small Python client.

About token scopes and 403s. Reading files, frames, and comments works on any plan. Variables (design tokens) typically require an Enterprise org, and Dev Resources require Dev Mode / the file_dev_resources:read scope — without them, those two calls return 403. Generate your PAT with the scopes you need before you start, not after the first failure.

Setup, in order

Section titled “Setup, in order”

Do these once, top to bottom:

  1. Install the figma pack — user scope by default, so your token stays yours: 
    # <catalogue> is your catalogue URI: a local clone path or a git+https://… URL.
    agentbundle install --pack figma --scope user <catalogue>
  2. Install the Python dependencies — python -m pip install credbroker httpx (httpx is the HTTP client; credbroker resolves your token — step 3’s pack also ships credbroker as a fallback, so this pip install just guarantees it’s importable).
  3. Install the credentials capability — the credential-brokers pack. It ships the credbroker resolver and the interactive credential-setup skill.
  4. Create your Figma PAT (see Before you install above).
  5. Complete credential setup interactively — tell your agent “set up credentials”. The credential-setup skill prompts you for the token and stores it in your OS keychain (or a 0600 dotfile). This step is yours, not the agent’s: the agent never types your token for you, and secrets never go on the command line or into the repo.
  6. Check the connection — ask your agent to “check my Figma connection”. It runs a reachability probe (the agent does this) and confirms your account is visible with no auth error.

Usage

Section titled “Usage”

Give your agent a normal Figma share URL and a frame name. For example:

  • “Read the structure of this Figma file and list its pages and top-level frames: <file-url>.”
  • “Render the ‘Login’ frame of <file-url> as a PNG.”
  • “Convert the FigJam connector graph in <file-url> to a Mermaid diagram.”

The result is a plain answer in chat (structure, comments) or a local image file (renders) — with the render receipt above so you can find it and confirm nothing changed on Figma.

→ Go deeper: the figma guides.