Framelink wins without Code Connect — 25% smaller, framework-agnostic output. Official wins with Code Connect — maps components to your codebase. CTO Guide and LogRocket independently confirm this split.
UX / UI
Five lanes: (A) read-only Figma context (Official #1, Framelink #2), (B) bidirectional write-access (Console MCP #1, Grab #2, figma-use #3), (C) alternative platforms (Penpot, Excalidraw), (D) specialized design-to-code agents (Kombai — 75–80% fidelity), (E) AI-native design creation (Google Stitch — Figma stock -8.8%, Onlook). Uber uSpec remains strongest enterprise validation. Google Stitch is provisional (2 days old).
13
Ranked
15
Signals
Current ranking
Best for: Teams with Figma Professional/Enterprise and Code Connect configured — the 'batteries included' choice
Highest trust: zero CVEs, triple AI partnership (Copilot, Claude Code, Codex). Bidirectional since Mar 6 2026. Code Connect (341K npm/week) maps components to your codebase. 14 tools, 14+ clients. 85–90% accuracy on well-structured files (CTO Guide).
Best for: Individual developers, free-tier Figma users, and teams with custom codebases where descriptive metadata is more useful than prescriptive code
Highest actual install volume in the category: 38.6K npm/week, 13.8K stars, 43.6K PulseMCP weekly visitors. 25 contributors. 25% smaller output than Official (CTO Guide). Works on any Figma account — no Dev Mode required.
Best for: Design system teams managing variables, tokens, specs, and multi-platform documentation at scale
Uber uSpec production usage — strongest enterprise validation in category (thousands of engineers, 7 stacks, accessibility in <2 min). 57+ tools (11 variable/token), 17.4K npm/week. @UberEng Twitter confirmed deployment.
Best for: Cursor users wanting bidirectional Figma interaction without Dev Mode costs
Cursor marketplace curation (alongside AWS, Stripe, Linear) — independent enterprise signal. 6,505 stars, 9.3K PulseMCP weekly visitors. Free-tier compatible via Figma plugin API. Product designer testimonial: 'nearly pixel-perfect flow in five minutes.'
Best for: Power users who want CLI-based Figma control and are comfortable with a single-maintainer dependency
Strongest HN signal in write-access lane (115pt, 37 comments). CDP-based, 100+ commands, 'JSX importing ~100x faster than any plugin API import' (HN commenter). Works without Figma Pro.
Best for: React/Next.js teams wanting to skip Figma entirely and have designers work directly in code
24.9K stars (highest in category), YC-backed. HN: 408pts/81c — strongest HN signal in category. 'Cursor for Designers' — the design IS the code, changes become PRs.
Best for: Teams where pixel-perfect fidelity matters more than cost or transparency
Highest fidelity in controlled comparisons: 75–80% vs 65–70% for MCP+LLM (FreeCodeCamp). Refactoring time ~2 min vs ~15 min. 30+ framework support. Multiple independent comparisons consistent.
Best for: Greenfield projects where you want AI-generated designs from prompts/voice, not Figma-to-existing-code
Google-backed, free (350 gen/month). Mar 18 overhaul caused Figma stock -8.8%. Infinite canvas, voice-driven 'vibe design', React export, DESIGN.md. MCP server + SDK confirmed.
Best for: Claude Code / Claude Desktop users wanting write-access on free Figma plans
Claude-specific with DXT one-click installer. 524 stars. Free Figma plan support. WebSocket-based architecture (port 3055) with Figma Plugin bridge.
Best for: Diagramming and whiteboarding — AI-driven visual design outside Figma
3.4K stars, official Excalidraw org backing, 26 tools with canvas inspection and screenshot capabilities. Compatible with Claude, ChatGPT, VS Code, Goose, Cursor.
Best for: Solo devs wanting free, code-native design-to-code without Figma
HTML/CSS-native canvas — 'the design is already code' (SFAI Labs). 24 MCP tools, full bidirectional. Abduzeedo covered GPU shader capabilities.
Best for: Design-system-forward experiments with agent guardrails
400 stars in 19 days — fastest growth rate in category. Only tool publishing model-specific design quality benchmarks. Harness engineering is novel.
Best for: Teams committed to open-source tooling or already on Penpot
Only open-source design platform with official MCP. Platform has 44.9K stars. MCP merged into main Penpot repo (Feb 2026). Smashing Magazine coverage.
Head to head
Console MCP has Uber production evidence, 25x npm downloads, and 3x the tool surface. Grab has more stars but Console has stronger enterprise validation. Grab wins on accessibility for individual devs and free-plan users.
The official path wins on trust and team adoption. Figma-use wins on direct write access and builder energy. Official is safer; Figma-use is more capable for power users.
Both offer official-lane trust. Figma MCP is the broader ecosystem play. Paper MCP is a narrower direct-write wedge outside Figma entirely. Different tools for different design surfaces.
Kombai wins on fidelity (75–80% vs 65–70%, FreeCodeCamp) and refactoring time (~2 min vs ~15 min). Framelink wins on transparency (open-source), cost (free), and ecosystem flexibility (any LLM). Different mechanisms: proprietary engine vs open MCP+LLM pipeline.
Both in AI-native design creation lane. Stitch is Google-backed with MCP support and AI generation from prompts. Onlook is open-source with 24.9K stars but declining activity. Stitch is greenfield design; Onlook is design-in-existing-code. Different starting points.
Public signals
Strongest enterprise validation in the entire category. Publicly traded company documenting production usage — automated specs across 7 stacks, accessibility in under 2 minutes. All local — no data leaves network.
Key independent comparison by Alex Bobes (CTO, Extremoo). Confirms Framelink cleaner without Code Connect, Official wins with Code Connect. Framework for the lane split.
Figma Official MCP gains bidirectional capability — generate_figma_design sends rendered UI back to Figma as editable frames. Transforms Official from read-only to full design loop. Currently the ONLY option with first-party bidirectional support.
RCE via command injection (CVSS 7.5) in Framelink <v0.6.3. Patched Sep 29, 2025. Security press recommends migrating to Official Figma MCP as the migration path. Users on old versions remain exposed. ⚠️ Enforce ≥v0.6.3 — check your lockfile.
Fastest-growing contender by download trajectory. Hockey-stick from 127 npm/week in November to 17,356 as of 2026-03-10 (API direct). If sustained, will challenge Framelink’s 38.6K/wk.
Independent comparison validates the simplicity-vs-depth trade-off. Confirms Framelink’s descriptive JSON vs Official’s prescriptive output. Third-party evidence supporting the lane split.
Highest trust in the category because it is the provider spelling out the workflow split directly.
The repo itself is a useful trust signal because it shows Figma treating this as a maintained, evolving workflow instead of a one-off demo.
Highest HN engagement for any Figma MCP (115 points, 37 comments). 8x the Official MCP’s launch (14 pts). Community validated CLI-driven write-access approach.
Onlook hit 408 HN points (May 2025) with a second hit at 227 points (Aug 2024) — highest HN validation in the UX/UI category. 24.9K stars but activity declining (last meaningful commit 2026-01-21). Monitor for stall.
Creator comparison — Official MCP is read-only window, Console MCP is full access. 57+ tools vs 14, WebSocket real-time awareness, variable management where Official has zero. Self-reported but objectively verifiable.
Google Stitch's Mar 18 overhaul caused Figma stock to drop 8.8% in a single day (-35% YTD). Market is pricing this as a serious threat to Figma's moat. MCP server + SDK confirmed — connects to Gemini CLI, Claude Code, Cursor. Provisional — zero usage evidence.
Controlled comparison showing specialized Figma-to-code engine (Kombai) outperforming MCP+LLM pipeline on fidelity. Not an MCP — proprietary Figma plugin. Indicates a ceiling on what generic MCP approaches can achieve.
Uber's official engineering Twitter account confirmed the Figma Console MCP deployment described in the Uber Engineering Blog. Cross-platform confirmation strengthens the enterprise validation signal.
Framelink has 43,600 PulseMCP weekly visitors — highest in the UX/UI category. Grab (cursor-talk-to-figma) at 9,300. Independent traffic signal confirming Framelink's adoption dominance in the read-only lane.
What changes this
Framelink loses #1 if: it remains read-only while bidirectional becomes table stakes, OR a second security incident hits without prompt patching, OR Graham Lipsman stops maintaining (bus factor = 1). Oct 2025–Jan 2026 npm decline (224K → 63K) shows it's not immune to competition.
Official overtakes Framelink if: Figma drops the Pro paywall for meaningful usage (>6 calls/month on free), OR Code Connect adoption reaches critical mass. Triple partnership gives distribution to eventually overtake — but pricing friction must drop first.
Console MCP rises to #2 if: a second major enterprise case study emerges (Uber alone is strong but single-source), OR npm growth continues hockey-stick trajectory past Framelink's 38.6K/wk.
If Grab launches enterprise features (variable management, cloud mode), it closes the gap with Console MCP and could reclaim #3.
Vibma enters ranked list if: independent HN/review validation appears within 2 weeks. At current growth rate (400 stars in 19 days), it will overtake figma-use on stars alone — but stars without independent evidence aren't sufficient.
If figma-use CDP architecture breaks permanently on Figma 126+, it drops below the cut line. Platform-dependent fragility is existential.
The entire Figma MCP category becomes less relevant if: AI-native design generation tools (Google Stitch, v0, Banani, Kombai) mature to the point where starting from a Figma file is no longer the default workflow.
If Google Stitch gains real adoption evidence, it could displace Figma as the starting point for AI-assisted design workflows. Watch for usage data beyond launch press.
If Kombai launches an MCP server, it collapses Lane D into Lane A/B and could challenge both Official and Framelink on fidelity.
If Onlook commits don't resume by Q2 2026, consider archiving. Activity has been declining for ~2 months.
If Anima MCP gets one independent review, it moves to ranked list — 1.5M plugin users is a large base waiting for MCP validation.
If Console MCP gets a second enterprise case study, it would cement Lane B dominance beyond any doubt.
If Figma Official makes Dev Mode free for MCP, Framelink loses its primary advantage and Official becomes clear #1 across Lane A.
If Excalidraw MCP hits 10K+ stars, it forces a category scope decision. At that scale, excluding it would be editorial malpractice.
If Paper.design exits alpha, it moves above the cut line immediately. HTML-native architecture is genuinely better for AI-first workflows.