b52d7e2651
* fix(app): record MCP and CLI structural edits as undo steps The automation bridge ran non-atomic tools, render, and eval without an undo entry, so Edit > Undo could not revert layers an MCP client or the CLI created, deleted, or rearranged. Snapshot the page around these edits as the AI chat does, and skip the entry when nothing changed so read-only scripts leave the history alone. * feat(app): activate documents, undo, redo, and change settings over automation Add activate_document, undo, redo, get_settings, and update_settings to the app's automation bridge. Settings cover appearance, snapping, canvas rendering, recovery, and chat preferences, validated with Valibot and applied through their owning stores; credentials, models, MCP connections, storage, and tool access stay out of reach. * feat(mcp): expose document activation, history, and settings tools * feat(cli): manage documents, history, settings, and tools in the running app Turn documents into a command group (list, open, new, save, close, activate), add undo, redo, and settings get/set, and add tool list/describe/call so every MCP tool runs from the shell, against the running app or headlessly on a file. * docs: document app control from the CLI and MCP * fix: never prompt in the app from automation closes and saves close_file opened the app's Save changes dialog, which an agent cannot answer: the call timed out and the dialog stayed open. It now fails on unsaved changes unless the caller passes unsaved "save" or "discard" (CLI --save or --discard). save_file and new_document no longer open a Save dialog for a document that was never saved, report a failed save as an error, and leave the document untouched when the path is refused. * docs: describe non-interactive close and save * fix: address review findings in app automation Keep a document's source when a save to a new path fails, report vector-edit undo and redo no-ops as unapplied, echo only the applied patch from update_settings so writing cannot read settings, reject tool call --write/--output without a file, and stop settings get from following inherited keys. * fix(app): record render undo on the page that receives the layers A render into a parent on another page was snapshotted against the target page, so undo left the new layers in place. Snapshot the page that contains the parent instead, and document that eval edits made after switching pages stay outside the undo step. * feat(app): limit automation undo to its own steps and expose design check settings The undo history is shared with the person in the editor, so an agent's undo could revert the user's last edit. Automation undo and redo now act only on steps made through the bridge, and only while they are newest; otherwise they fail and leave the history alone. Vector edit mode's session history is off limits entirely. Settings automation also covers the design check preferences that landed on master.
5.2 KiB
| title | description |
|---|---|
| Controlling the App | Open, save, switch, and close documents, undo and redo, change settings, and call any editor tool in the running app from the CLI. |
Controlling the App
While the desktop app is running, the CLI drives it over the same automation bridge as the MCP server. Every command on this page except headless tool call needs the running app.
Documents
Each open tab is a document with a stable ID. List them, then pass --document-id (and --page-id) to aim a command at a background tab without switching to it:
openpencil documents list --json
openpencil tree --document-id tab-123 --page-id 0:1
Open, create, save, and close documents:
openpencil documents open designs/landing.fig # opens in a new tab
openpencil documents new --path designs/draft.fig # empty document, saved to the path
openpencil documents save --document-id tab-123
openpencil documents save --document-id tab-123 --path designs/copy.fig
openpencil documents close --document-id tab-123
openpencil documents close --document-id tab-123 --save # or --discard
Relative paths resolve against the shell's working directory. These commands never open a dialog in the app, because nobody may be there to answer it. Closing a document with unsaved changes fails unless you pass --save or --discard, and saving a document that has never been saved needs --path. Each command prints the document and page it acted on; with --json it prints { "result", "target" }, so a script can read the new document's ID from target.documentId.
To bring a tab to the front, optionally on a specific page:
openpencil documents activate tab-123 --page-id 0:4
Undo and redo
Step back through changes made through the CLI and MCP, like Edit → Undo and Edit → Redo:
openpencil undo --document-id tab-123
openpencil redo --document-id tab-123 --json
{
"result": { "applied": true, "label": "Set opacity", "scope": "document" },
"target": { "documentId": "tab-123", "documentName": "Landing", "pageId": "0:1", "pageName": "Page 1" }
}
When there is nothing to undo or redo, the command says so and result.applied is false.
The editor has one history, shared by you and every automation client, so Edit → Undo in the app steps back through any change, newest first. Automation is narrower: undo and redo only act on a step that the CLI or MCP made, and only while it is the newest one. If the newest change was made in the editor, the command fails and leaves it alone, so an agent can't revert your work. They also fail while the document is in vector edit mode, whose history belongs to the editing session.
Each CLI or MCP command that edits the document is one undo step. An eval script is recorded against its target page: if it switches figma.currentPage and edits another page, those edits are not part of the undo step.
Settings
Read and change editor settings by dotted key:
openpencil settings get
openpencil settings get appearance.theme
openpencil settings set appearance.theme light
openpencil settings set editing.snapping.pixelGrid false
openpencil settings set chat.maxAgentSteps 100
Values are read as JSON when they parse (false, 100, "auto") and as plain strings otherwise. Invalid keys and values are rejected without changing anything.
| Key | Values |
|---|---|
appearance.theme |
dark, light, auto |
appearance.language |
en, de, es, fr, it, ja, pl, ru, zh-CN |
appearance.animations |
system, off |
editing.snapping.geometry |
true, false |
editing.snapping.objects |
true, false |
editing.snapping.pixelGrid |
true, false |
rendering.canvasMode |
retained, tiled (applies after a reload) |
recovery.enabled |
true, false |
chat.reasoningDisplay |
collapsed, while-thinking, expanded |
chat.maxAgentSteps |
integer 1–1000 |
chat.changePreviewSize |
off, small, medium, large |
designCheck.showOnCanvas |
true, false |
designCheck.preset |
recommended, strict, accessibility |
designCheck.disabledRules |
JSON array of rule IDs, e.g. ["no-default-names"]; see openpencil lint --list-rules |
Credentials, AI models, MCP connections, storage, and tool access are deliberately not available here: an automation client can't read secrets or grant itself access.
Calling tools
Every editor tool the MCP server exposes is also available from the CLI. List them, inspect a tool's arguments as JSON Schema, and call it with a JSON object:
openpencil tool list
openpencil tool describe set_fill
openpencil tool call set_fill --args '{"id":"0:5","color":"#2563eb"}'
openpencil tool call set_fill --document-id tab-123 --args-file fill.json
echo '{"name":"Icons"}' | openpencil tool call create_page --args-file -
Pass a file to run a tool headlessly instead. Changes to a file are kept only with --write (back to the input) or --output:
openpencil tool call create_page design.fig --args '{"name":"Icons"}' --write
openpencil tool call get_page_tree design.fig --json
For multi-step edits, eval runs a whole script against the same Figma-compatible API.