TalkToFigma Desktop MCP Tool Usage Deep Dive
This report focuses on successful mcp_tool_call usage in TalkToFigma Desktop. The goal is to explain what people actually used the MCP tools for: not just whether they connected, but which design tasks they repeatedly automated.
The strongest pattern is that users repeatedly used MCP for real Figma editing work: creating text, applying colors, creating frames, rounding corners, editing text content, creating rectangles, inspecting nodes, moving layers, and exporting nodes as images.
Retrospective takeaway: users did not merely try the connection. They repeatedly used MCP to produce and refine design objects.
What People Automated
The top successful tools group into six design purposes. The largest three are content/text editing, structure/layout building, and visual styling. Together they describe a practical design-production loop: create content, build layout structure, and style the result.
Successful MCP Calls By Design Purpose
Event count, grouped by inferred purposeMost Repeated Tools
The highest-volume successful tools are the everyday design actions: create text, set fill color, create frames, round corners, edit text, and create rectangles. These are not incidental diagnostics; they are concrete editing operations.
Top Successful MCP Tools
Ranked by successful event countBreadth And Repeat Intensity
A quadrant view is clearer than a scatterplot for this report. It separates tools that were broadly adopted from tools that were niche, and tools that were repeated heavily from tools that were used more lightly.
Thresholds: Broad means at least 300 active users. Repeat-heavy means at least 25 successful calls per active user.
MCP Tool Usage Quadrants
Successful events by usage quadrant| Quadrant | Successful events | Tools | Representative tools | Interpretation |
|---|---|---|---|---|
| Broad + repeat-heavy | 442,564 | 11 | create_text, set_fill_color, create_frame, set_corner_radius, set_text_content |
Core workflow: many users repeatedly created content, styled layers, built frames/shapes, and inspected nodes. |
| Broad + context/light repeat | 39,670 | 10 | get_document_info, get_selection, get_active_channels, connection_diagnostics |
Widely used context and diagnostic actions; important for grounding, but less repetitive per user. |
| Niche + repeat-heavy | 35,548 | 7 | set_layout_sizing, clone_node, set_item_spacing, set_axis_align, set_auto_layout |
Advanced automation: fewer users, but repeated heavily once adopted. |
| Niche + light repeat | 9,031 | 4 | set_layout_mode, set_padding, get_nodes_info, delete_multiple_nodes |
Secondary workflows with narrower reach and lower repeat intensity. |
Tool Usage Interpretation
The table below keeps the strongest individual rows from the report. It includes inferred purpose, successful events, active users, and calls per user.
| Tool | Purpose group | Likely purpose | Successful events | Active users | Calls / user | Retrospective interpretation |
|---|---|---|---|---|---|---|
create_text |
Content | Create new text layers | 105,643 | 613 | 172.3 | The strongest signal: users repeatedly generated text objects, likely for AI-assisted layout drafts and content-heavy UI screens. |
set_fill_color |
Visual styling | Apply fill colors | 92,561 | 604 | 153.2 | Color application was one of the most repeated edits, suggesting users used MCP to style generated or selected layers. |
create_frame |
Structure/layout | Create frames and containers | 52,509 | 613 | 85.7 | Users were not only editing existing nodes; they were constructing screen and layout structure. |
set_corner_radius |
Visual styling | Round corners | 38,053 | 504 | 75.5 | High repeat usage points to systematic styling of buttons, cards, rectangles, and UI components. |
set_text_content |
Content | Edit text content | 37,845 | 455 | 83.2 | Users iterated on copy after text layers existed, which fits real UI and content production. |
create_rectangle |
Structure/layout | Create shape primitives | 34,779 | 473 | 73.5 | Basic shape creation was heavily used, likely for UI blocks, backgrounds, controls, and placeholders. |
get_node_info |
Context | Inspect selected or referenced nodes | 32,736 | 842 | 38.9 | The largest active-user base among editing-adjacent tools; agents needed context before editing. |
move_node |
Structure/layout | Reposition nodes | 13,859 | 324 | 42.8 | Layout adjustment was a real repeated workflow, not just object creation. |
export_node_as_image |
Asset export | Export selected nodes as images | 13,536 | 455 | 29.7 | Users also used MCP to inspect, share, or render design output, not only modify Figma files. |
set_layout_sizing |
Structure/layout | Set auto-layout sizing behavior | 9,018 | 167 | 54.0 | Lower adoption but high repeat intensity, suggesting more advanced layout automation. |
Small-Sample Caveat
Some tools have very high repeat intensity but a tiny active-user base. They are interesting, but should not be presented as broad user behavior.
| Tool | Purpose group | Successful events | Active users | Calls / user | Caveat |
|---|---|---|---|---|---|
set_auto_layout |
Structure/layout | 3,208 | 5 | 641.6 | Very high repeat intensity, but too few users for a broad behavioral claim. |
rename_node |
Cleanup | 1,329 | 5 | 265.8 | Likely heavy-user automation or testing; mention only with sample-size caveat. |
Source Details
- Original artifact URL
- https://mcp-server-dataanalyticswidgets-d90d6b74b2c37858.web-sandbox.oaiusercontent.com/?app=skybridge
- Generated at
- 2026-07-01T00:00:00+09:00
- Local raw sources
/Users/jude.park/Sites/Retro/talk-to-figma-desktop-ga4-raw-data.json/Users/jude.park/Sites/Retro/talk-to-figma-desktop-ga4-raw-data.md- GA4 property
properties/499789310· Figma MCP Agent- Date range
- 2025-07-01 to 2026-06-30 · observed effective range starts 2025-08
- Metric definitions
eventCountandactiveUsersare GA4 metrics.callsPerUseris calculated as successful tool events divided by active users for that tool.- Quadrant thresholds
- Broad:
activeUsers >= 300. Repeat-heavy:callsPerUser >= 25. - Export notes
- This is a static portable HTML export. It intentionally omits the Data Analytics interactive top bar, refresh controls, export controls, and app-only controls.