| name | figma-edit |
| description | Use when reading or editing a Figma file through the figma-edit-mcp server — inspecting nodes; creating or modifying shapes, frames, text, styles, variables, components, instances, annotations, or prototype reactions; and recovering from the server's structured errors. Explains the plugin's hard constraints (scope locking, name verification, batch validation) and how to pick the right tool. |
figma-edit
figma-edit-mcp lets an agent read and edit a live Figma document through a plugin that enforces safety constraints at execution time. Two things to internalize before acting:
- Discover before acting. Every workflow that touches a node starts with a read (
page_info → node_info). IDs and names the user gives you may be stale; verify them.
- The plugin can refuse you. Hard constraints (scope, name verification, per-item batch checks) are checked inside Figma and return structured error codes. You cannot bypass them — plan calls to comply up front.
- Explicit variable scopes. When creating a new variable, you must explicitly provide its
scopes.
References
Load the one you need:
- references/constraints.md — the hard rules the plugin enforces (scope locking, name verification, no implicit selection, batch validation, URL IDs) and what to do when a constraint forbids the request.
- references/error-playbook.md — every structured error code and its correct recovery.
- references/workflows.md — the discover-before-acting pattern and canonical recipes (find-and-update text, create-in-frame, edit-by-URL, bulk text replacement).
- references/tool-selection.md — choosing among overlapping tools,
node_info field selection (fast vs slow), batch vs single, and streaming progress.
The same content is available over the MCP connection as resources under figma-edit://guide/*.