Ligma: a better Figma MCP tool / CLI command for coding agents

CLI alternative to the Figma MCP. Fetches a Figma node and emits a deterministic, token-efficient artifact bundle optimized for code agents.

The problem it sets out to solve: helping agents make sense of the things they need to know to implement a design. Figma json is geared towards accurately recording document structure, not revealing intent and guiding implemention.

Files are downloaded to disk instead of dumped into the agent's context window (./artifact by default, but can be specified), and can list / download additional assets (as png or svg).

Made with spec-driver.

Install

npm install -g ligma-mcp

Or from source:

git clone https://github.com/davidlee/ligma && cd ligma
pnpm install && pnpm build && pnpm link --global

Requires Node >= 18.

MCP server setup (Claude Code)

claude mcp add ligma -- ligma mcp

Requires FIGMA_TOKEN in your shell environment. Alternatively, pass it explicitly:

{
  "mcpServers": {
    "ligma": {
      "command": "ligma",
      "args": ["mcp"],
      "env": {
        "FIGMA_TOKEN": "your-figma-token"
      }
    }
  }
}

6 tools: figma_get_node, figma_get_outline, figma_get_render, figma_list_assets, figma_get_asset, figma_get_assets.

Usage

export FIGMA_TOKEN="<your-figma-personal-access-token>"

ligma <figma-url>

Or pass the token inline:

ligma <figma-url> --token <token>

Note: it takes a hot minute and there's no feedback during download. Little spinner coming soon (ironic, eh?).

Subcommands

list-assets

List detected export targets as JSON:

ligma list-assets <figma-url>

Outputs a JSON array of { nodeId, name, format, reason } to stdout.

| Flag | Default | Description | |------|---------|-------------| | -t, --token <token> | $FIGMA_TOKEN | Figma personal access token | | -d, --depth <n> | 2 | Node tree depth | | --max-assets <n> | 20 | Maximum assets to list | | --no-cache | | Disable fetch caching | | --cache-directory <path> | .cache/figma-fetch | Cache directory path |

get-asset

Fetch a single asset by node ID and write to disk:

ligma get-asset <figma-url> <node-id>

Writes the asset file and prints the absolute path to stdout.

| Flag | Default | Description | |------|---------|-------------| | -t, --token <token> | $FIGMA_TOKEN | Figma personal access token | | -o, --out <dir> | ./artifacts | Output directory | | -f, --format <fmt> | png | Image format (png or svg) | | --no-cache | | Disable fetch caching | | --cache-directory <path> | .cache/figma-fetch | Cache directory path |

Options

| Flag | Default | Description | |------|---------|-------------| | -t, --token <token> | $FIGMA_TOKEN | Figma personal access token | | -o, --out <dir> | ./artifacts | Output directory | | -f, --format <fmt> | png | Image format (png or svg) | | -s, --scale <n> | 2 | Image scale (0.01 – 4.0) | | -d, --depth <n> | 2 | Node tree depth | | --include-hidden | false | Include hidden nodes in outline and context | | --no-expand | | Disable selective expansion | | --max-expand <n> | 10 | Maximum expansion targets | | --expand-depth <n> | 2 | Depth for expansion refetches | | --max-assets <n> | 20 | Maximum assets to export | | --asset-format <fmt> | auto | Asset export format (auto, png, svg) | | --no-cache | | Disable fetch caching | | --cache-directory <path> | .cache/figma-fetch | Cache directory path |

Project documentation

• Technical brief
• PROD-001 — Figma fetch and normalization pipeline

Design revisions

| DR | Delta | Description | |----|-------|-------------| | DR-001 | DE-001 | PROD-001 delivery planning | | DR-002 | DE-002 | Project scaffold and Figma client | | DR-003 | DE-003 | Core normalization engine | | DR-004 | DE-004 | Semantic inference, tokens, and assets | | DR-005 | DE-005 | Outline and context generation | | DR-006 | DE-006 | Selective expansion and caching | | DR-007 | DE-007 | Adopt strict lint config | | DR-008 | DE-008 | Backlog fixes: individualStrokeWeights and interaction extraction | | DR-009 | DE-009 | Asset export pipeline | | DR-010 | DE-010 | Minimal MCP transport layer |