figcraft/docs/cli.md

figcraft-cli

FigCraft without a window: inspect, export, convert and render documents, run command scripts, list commands, serve MCP and measure performance. Build it with cargo build --release -p figcraft-cli (binary target/release/figcraft-cli).

Bad arguments print usage and exit with code 2; failures exit with 1. The CLI never panics on input.

Fonts

Text lays out and renders with the same fonts as the desktop app: the fonts built into FigCraft plus the installed system fonts. The system fonts are indexed once per run, the first time text is laid out (about 60 ms warm on a Windows machine with ~900 font files; a run without text skips it). Every subcommand takes --no-system-fonts (or set FIGCRAFT_NO_SYSTEM_FONTS=1, which the desktop app honours too) to use only the built-in fonts, for renders that come out the same on every machine, e.g. in CI.

On Linux and other Unix systems except macOS, user fonts are read from $XDG_DATA_HOME/fonts, or $HOME/.local/share/fonts when XDG_DATA_HOME is unset, empty or relative. The legacy $HOME/.fonts directory is also scanned.

The collaboration server is a binary of its own, figcraft-server (cargo build --release -p figcraft-server): it keeps files with share links and serves their live sessions, and its key, files, links, reports, block-ip and check subcommands manage them (a server can be open to the public: server.md). See server.md.

Subcommands

info <file> [--json]

Pages, node counts, top-level frames (id, type, name, size), style/variable/image counts. Repairs of a damaged file and import warnings go to stderr as note: lines (notes with --json). Fonts the file names that aren't installed are listed as missing font lines (missingFonts: [{family, layers}] with --json); their text renders in a fallback font.

Every subcommand that reads a file also takes the public REST API file JSON (the response of GET /v1/files/:key or of its nodes endpoint, written to a .json file): it is imported on the fly. Image fills come from images/<imageRef>.png (or .jpg, …) next to the JSON when present, else a grey placeholder. .fig files are read the same way (best effort, images from the file itself); info lists what wasn't imported as note: lines.

$ figcraft-cli info app.figcraft
app.figcraft — App (412 nodes, 6 styles, 12 variables, 3 images)
page 1 "Screens": 380 nodes, 4 top-level
  2 FRAME "Home" 393.0×852.0
  …

export <file> -o <out> [--format png|jpg|svg|webp|pdf|fig] [--scale N] [--page NAME] [--node ID|NAME] [--bleed LENGTH] [--crop-marks] [--bleed-marks] [--cmyk]

• <out> is a directory: one file per top-level frame of the page (default the first page), named after the frame (Home.png, Settings.png…); with no frames, one file per top-level layer.
• <out> has an image extension: one file with the given --node, or the whole page composited.
• --node takes a layer id (12 or 0:12) or a layer name.
• --format fig (or an <out> ending in .fig) writes the whole file as .fig, experimental: written from public format notes, including variables with their modes and bindings and shared styles with their layer links. Checked by opening it again in FigCraft and, on the example documents and small test files, in other software; other apps may not open every file. What was left out (comments, Dev Mode data, some bindings and settings…) is printed as note: lines; docs/file-format-parity.md lists what was checked and what is unchecked.
• The format comes from --format, else <out>'s extension, else png. --scale (default 1, max 64) applies to raster formats.
• SVG/PDF exports with visible descendants more than 256 levels below an exported root fail with an error instead of writing an incomplete file. Select an inner layer with --node or reduce nesting. An existing destination file is kept on this failure.
• A .pdf <out> without --node holds the page's top-level frames, one PDF page each in reading order, each the frame's size in points (a page without frames goes on one page). Shapes, text and gradients stay vector, text is real (selectable, with subset fonts) and its links clickable; shadows and blurs are images.
• PDF print options: --bleed (points, or a length like 3mm or 0.125in, up to 2 in) grows each page by the bleed, which shows what lies past the frame's edge (its fills reach out to it, its children are clipped there instead); --crop-marks and --bleed-marks draw hairline marks outside the bleed; the page boxes are set (MediaBox the sheet, BleedBox, TrimBox the frame). --cmyk writes CMYK colour, images included, converted from sRGB with the naive formula K = 1 − max(R, G, B), C = (1 − R − K) / (1 − K) (M and Y alike): DeviceCMYK without an output intent (no CMYK profile ships with FigCraft), so proof against your printer's profile.

figcraft-cli export app.figcraft -o exports/ --scale 2
figcraft-cli export app.figcraft -o hero.svg --node "Hero"
figcraft-cli export app.figcraft -o screens.pdf --page "Screens"
figcraft-cli export app.figcraft -o poster.pdf --node "Poster" --bleed 3mm --crop-marks --cmyk
figcraft-cli export app.figcraft -o handoff.fig

convert <in> <out>

By extension. Inputs: .figcraft, .json (FigCraft or REST API file JSON), .fig (best effort), .svg (an editable frame: groups, shapes, text, gradients, masks and effects; info lists what it left out as note: lines), .png/.jpg/.webp/.gif/.tif/.bmp (an image layer). Outputs: .figcraft, .json (FigCraft JSON, pretty-printed), .svg, .png/.jpg/.webp (the current page rendered at 1x with its background), .pdf (the current page's frames, one page each), .fig (the whole file, experimental, as export --format fig).

render <file> -o <out.png> [--scale N] [--page NAME]

The whole page (default the first) with its background, as png, jpg or webp. Large pages are scaled down to fit 16384 px.

run <script.json|script.jsonl> [--in <file>] [--out <file>]

Runs commands in a headless session (a new document, or --in), printing one JSON line per step:

{"command":"frame.create","ok":true,"result":{"id":2},"step":0}

The whole script is one undo step, and it is all-or-nothing, exactly like running the same commands through command.batch (the MCP tool batch, the control channel, plugins): {"$step": n} in a step's params is the id in an earlier step's result, so a step can target a layer an earlier step of the same script created:

[
  {"command": "node.create", "params": {"type": "frame", "name": "Card", "width": 200, "height": 100}},
  {"command": "node.create", "params": {"type": "text", "parent": {"$step": 0}, "text": "Hi"}}
]

The result lines are printed once the script is through, so none of them reports work a later step took back.

A failing step prints one line with "ok": false, the error and "rolledBack": true, writes nothing (--out is not saved, and the session's documents go back to how the script found them), and exits with code 1:

{"step":1,"command":"node.set","ok":false,"error":"no such node 99999","rolledBack":true}

--out saves the result by extension like convert.

Script formats: a JSON array of {"command": id, "params": {...}}, an object {"commands": [...]}, or JSON lines (one step per line; blank lines and // lines are skipped).

[
  {"command": "frame.create", "params": {"preset": "iPhone 16", "x": 0, "y": 0, "name": "Home"}},
  {"command": "node.create", "params": {"type": "text", "parent": 2, "x": 24, "y": 64, "text": "Welcome", "props": {"fontSize": 32}}},
  {"command": "layout.add", "params": {"ids": [2], "direction": "vertical"}}
]

Ids are assigned in order (a new document's first page is 1, so the first layer is 2), which lets a script refer to layers it created.

A script can start from a starter template instead of an empty document: template.list names them and file.newFromTemplate opens one as the document the following steps (and --out) act on.

echo '[{"command": "file.newFromTemplate", "params": {"template": "dashboard"}}]' > start.json
figcraft-cli run start.json --out dashboard.figcraft
figcraft-cli export dashboard.figcraft -o dashboard.png

Confining a run: --automation-read-root DIR / --automation-write-root DIR

info, export, convert, render and run take these two flags, with the same meaning they have on mcp: each is repeatable, each folder must exist, and together they confine every file that subcommand touches. Reads are allowed inside a read or write root, writes only inside a write root, and everything else is refused — so a run given only a read root cannot write at all. Without either flag a subcommand reads and writes as the user who started it.

# Renders in/doc.figcraft to out/ok.png and can touch nothing else.
figcraft-cli render in/doc.figcraft -o out/ok.png \
  --automation-read-root in --automation-write-root out

figcraft-cli render in/doc.figcraft -o ./escaped.png \
  --automation-read-root in --automation-write-root out
# error: automation path rejected: . is outside the folders that may be written: …/out

This is for driving the CLI from a script or an agent that should not reach the rest of the disk; the flags leave the output identical when nothing strays outside the roots.

commands [--json] [--filter TEXT]

Every command with its label, shortcut, undoable flag (*) and parameter docs; --json for machines.

mcp [--attach PORT] [--automation-read-root DIR]… [--automation-write-root DIR]…

Serve MCP on stdio: headless by default, or attached to a running app started with figcraft --control PORT. See mcp.md. The automation roots (each repeatable; existing folders) confine the files the agent's calls read (inside a read or write root) and write (inside a write root); everything else is refused (mcp.md › Confining file access). With --attach they are refused: start the app with them instead. The headless session can host, join and publish live sessions (collab.*: without a key on a server open to the public, with collab.serverInfo, collab.manageFile for the owner token and collab.report) and use plugin directories (plugin.directory.*); the one-shot run can't (no network link).

plugin publish <folder> --server ws://host:port/[?key=KEY]

Publishes a plugin folder (figcraft-plugin.json, the module its main names, an optional assets/ folder) to a figcraft-server's plugin directory as a new version, and prints its name, version, size and SHA-256. The server's key comes from the address or, better, from the FIGCRAFT_SERVER_KEY environment variable (it then stays out of the shell history and process lists); it is never printed. The bundle is checked before it is sent and again by the server. See plugins.md.

perf [--layers N[,N...]] [--instances N[,N...]]

Builds synthetic documents of each size (default 1000,10000 layers: screens with an auto layout header, a card group, button instances and a grid of shapes and text) and prints a Markdown table of the hot paths: build + normalise, save/load, 1440×900 renders at 100% and zoomed to fit, hit testing and hover, single-layer edits through the engine (fill, move, text in auto layout, a main component with all its instances), undo/redo, a drag preview frame, normalize after a one-node edit, document.inspect and select-all + move. Use a release build for meaningful numbers; cargo xtask bench runs it at 1k/10k/50k layers together with the UI frame benchmarks.

--layers picks the document sizes (each value is clamped to 10 to 2,000,000 layers). --instances picks the component workloads (default 2000): for each count the command builds a synthetic design system (about 300 components, variables and styles, nested four deep) with that many instances that carry overrides at every level, in one file and as a library in another. It then prints a second table with the build time and edits of a deep main component (icon colour, glyph size, an added layer) and of the 72 button mains, where all copies follow. Zeros are dropped and counts above 100,000 are capped to 100,000; --instances 0 skips this table.