No user-submitted related servers found.
Scored across 140 tools
Most tools have a sharply scoped purpose and the descriptions cross-reference siblings explicitly (e.g. figma_set_fills vs figma_apply_style_to_node, figma_get_reactions vs figma_get_motion_context). But the set is so large that near-neighbours abound: six-plus read tools (get_node, get_nodes_info, get_design_context, get_metadata, search_nodes, scan_text_nodes, scan_nodes_by_types) and five export tools (export_node, save_screenshots, download_assets, export_frames_to_pdf, export_video) require reading long descriptions to choose correctly. An agent will frequently pick the wrong one on first pass.
Overwhelmingly a figma_verb_noun snake_case pattern (create_frame, set_fills, get_design_context, delete_variable), which is predictable and readable. Minor deviations exist: figma_batch, figma_savepoint, figma_sign_off, figma_stress_test, figma_status, figma_batch_rename_nodes mix noun-style names into the verb scheme, but none are ambiguous.
140 tools is far beyond any reasonable single-server surface and forces the agent to triage a huge catalogue before acting. Even with excellent descriptions, the cognitive load and misselection risk are extreme; many tools (create_rectangle, create_ellipse, create_polygon, create_star, create_line, create_vector) are thin wrappers that should be consolidated into one shape tool or a spec builder.
Coverage is exceptional for the domain: full node CRUD (create/clone/move/resize/rotate/reorder/reparent/group/boolean/flatten/delete), variables and styles end-to-end, components/variants/properties/slots, FigJam board primitives, Slides decks and transitions, motion keyframes and review, plus briefs, sign-off, savepoints and rollback. The only real gap is consolidation rather than missing capability.