{
  "schema_version": "2.0",
  "slug": "glips-figma-context-mcp",
  "name": "Framelink",
  "agent_url": "https://www.framelink.ai/",
  "category": "Coding",
  "run_id": "run-r4-v2-glips-figma-context-mcp-2026-08-27",
  "run_at": "2026-08-27T00:30:54.312Z",
  "editor": "Hlido Editor",
  "editorial_method": "public-surface-tier-1+editorial-narrative-v2",
  "methodology_version": "2026.05",
  "methodology_url": "/methodology/public-surface-tier-1/",
  "score": 80,
  "tier": "STEADY",
  "laddoo_score": 80,
  "confidence": "high",
  "hlido_opinion": {
    "headline": "A Figma-to-code MCP server that deliberately sends descriptive data rather than generated React — and publishes an honest, self-deprecating comparison against Figma’s own MCP.",
    "body": "Framelink solves a narrow problem well and explains its reasoning better than most tools in this category bother to. It pulls layouts, styles and components straight out of Figma into a coding agent's context over MCP, so the agent stops working from screenshots. The interesting decision is what it sends: descriptive JSON ('this element has a 1px border and 16px padding') rather than the prescriptive React and Tailwind that Figma's official MCP emits. The vendor's argument is that prescriptive output poisons the context — the model mimics auto-generated structure with arbitrary pixel values instead of using your components and conventions, so you end up refactoring the AI's output rather than using it. That is a real, specific failure mode, and the page shows the two outputs side by side rather than asserting the difference. Adoption backs the design: 15,000+ GitHub stars, and it advertises working with Cursor, Claude Code, VS Code, Codex and OpenCode. What earns Framelink credit beyond the engineering is the footnote under its own comparison table: the ~25% size advantage is marked '*Based on a few admittedly haphazard tests. I should probably do more.' A vendor that labels its own benchmark as weak is telling you where the evidence ends, and we would rather rate that than a confident unsourced number. The limits are the flip side of the focus: this is a single-purpose connector with no pricing or commercial-support story on the surface, the paid tier is a waitlist ('Special features coming soon. Sign up for early access'), and the size and quality claims remain the vendor's own haphazard tests, not an independent measurement.",
    "voice": "Hlido Editor",
    "as_of": "2026-08-27",
    "editor_signature_pending": true
  },
  "tier_rationale": "STEADY (80) because the design choice is technically specific and demonstrated side-by-side rather than asserted, adoption is substantial at 15,000+ stars across every major MCP host, and the vendor labels the weakness of its own benchmark in plain language — an honesty signal we weight. Held below VITAL because the headline ~25% advantage rests on the vendor's self-described 'haphazard' tests, there is no published pricing or support commitment, and the commercial tier is a waitlist rather than a product.",
  "what_it_does_well": [
    "Sends descriptive design data instead of prescriptive generated code, so the agent uses your components and conventions",
    "Preserves style names and represents nested components accurately rather than flattening them",
    "Publishes a direct side-by-side comparison against Figma’s official MCP instead of asserting superiority",
    "Explicitly labels its own benchmark as weak — \"a few admittedly haphazard tests\" — rather than presenting it as rigorous",
    "Broad host support: Cursor, Claude Code, VS Code, Codex, OpenCode and other MCP clients",
    "15,000+ GitHub stars indicates real adoption for a single-purpose connector"
  ],
  "what_it_fails_at": [
    "The ~25% output-size advantage is the vendor’s own self-described haphazard testing, not an independent measurement",
    "No published pricing, licence terms or support commitment on the reviewed surface",
    "Paid capability is a waitlist (\"Special features coming soon\") rather than a shipped tier",
    "Single-purpose by design — it does Figma-to-context and nothing else",
    "Output quality still depends entirely on the downstream agent; Framelink supplies context, not code"
  ],
  "best_for": [
    "Teams whose coding agent produces design-faithful markup but drifts from their own component conventions",
    "Anyone currently pasting Figma screenshots into a coding agent",
    "Design-system codebases where preserved style names matter more than generated markup"
  ],
  "not_recommended_for": [
    "Teams that want generated React/Tailwind handed to them directly — Figma’s own MCP is the prescriptive option",
    "Buyers who need a commercial support agreement or published SLA",
    "Non-Figma design workflows"
  ],
  "red_flags": [],
  "compared_to": [
    {
      "slug": "cursor",
      "verdict_diff": "Cursor is the agent host; Framelink is a context source that plugs into it over MCP. They are complementary rather than alternatives — Framelink improves what Cursor sees about a design, it does not replace the editor.",
      "preferred_for_axis": "context-source-vs-agent-host"
    },
    {
      "slug": "bgauryy-octocode-mcp",
      "verdict_diff": "Both are MCP context servers that feed a coding agent evidence instead of guesses. Octocode sources cited context from GitHub, npm and local code; Framelink sources it from Figma design files. Pick by where your missing context lives.",
      "preferred_for_axis": "design-context-vs-code-context"
    }
  ],
  "evidence_urls": [
    {
      "claim": "Pulls layouts, styles and components from Figma into an AI coding workflow over MCP",
      "source": "https://www.framelink.ai/ ('Framelink pulls layouts, styles, and components straight from Figma into your AI’s workflow.')",
      "tested_at": "2026-08-27",
      "verified": true
    },
    {
      "claim": "Sends descriptive JSON rather than prescriptive React/Tailwind; preserves style names; represents nested components accurately",
      "source": "https://www.framelink.ai/ (comparison table: 'Output format Descriptive JSON / Prescriptive React/Tailwind'; 'Style names Preserved / Lost'; 'Nested components Accurately represented / Flattened (misleading)')",
      "tested_at": "2026-08-27",
      "verified": true
    },
    {
      "claim": "The ~25% size advantage is the vendor’s own self-described haphazard testing",
      "source": "https://www.framelink.ai/ ('*Based on a few admittedly haphazard tests. I should probably do more.')",
      "tested_at": "2026-08-27",
      "verified": true
    },
    {
      "claim": "15,000+ GitHub stars; works with Cursor, Claude Code, VS Code, Codex, OpenCode",
      "source": "https://www.framelink.ai/ ('15,000+ stars'; 'Works with all agents: Cursor, Claude Code, VS Code, Codex, OpenCode')",
      "tested_at": "2026-08-27",
      "verified": true
    }
  ],
  "agent_relevance": {
    "has_api": false,
    "has_cli": false,
    "has_mcp": true,
    "has_webhook": false,
    "has_sdk": false,
    "behavioral_testable": true,
    "agent_integration_path": "Connect the Framelink MCP server to any MCP host (Cursor, Claude Code, VS Code, Codex, OpenCode). The agent then calls it to retrieve descriptive layout, style and component data for a Figma node, and generates code in whatever language the user asks for. It is a pure context provider — the agent does the generation.",
    "agent_friendly_score": 8
  },
  "marking_signal": {
    "not_applicable": true,
    "checked_at": "2026-08-27",
    "note": "Developer context server for design-to-code; it produces structured design data for an agent, not public-facing synthetic media, so machine-readable output marking is not an applicable obligation here."
  },
  "checklist": [
    {
      "id": "homepage_loads",
      "pass": true,
      "required": true,
      "tested_at": "2026-08-27"
    },
    {
      "id": "primary_value_prop",
      "pass": true,
      "required": true,
      "evidence": "'Turn Figma designs into pixel-perfect code in one shot'",
      "tested_at": "2026-08-27"
    },
    {
      "id": "cta_present",
      "pass": true,
      "required": true,
      "evidence": "'Quickstart' / 'Sign up for early access'",
      "tested_at": "2026-08-27"
    },
    {
      "id": "pricing_or_access",
      "pass": false,
      "required": true,
      "evidence": "No pricing published; paid capability is a waitlist — \"Special features coming soon. Sign up for early access.\"",
      "tested_at": "2026-08-27"
    },
    {
      "id": "competitive_claim_sourced",
      "pass": true,
      "required": true,
      "evidence": "Comparison table against Figma’s official MCP with a self-declared weak-evidence footnote",
      "tested_at": "2026-08-27"
    }
  ],
  "evidence_images": {
    "run_id": "run-a626c0a0cef117fa-www-framelink-ai",
    "base": "https://images.hlido.eu/reviews/glips-figma-context-mcp/run-a626c0a0cef117fa-www-framelink-ai",
    "files": [
      "home.png",
      "pagewaitlist.png",
      "pagedocs.png",
      "pagedocs_quickstart.png"
    ],
    "urls": [
      "https://images.hlido.eu/reviews/glips-figma-context-mcp/run-a626c0a0cef117fa-www-framelink-ai/home.png",
      "https://images.hlido.eu/reviews/glips-figma-context-mcp/run-a626c0a0cef117fa-www-framelink-ai/pagewaitlist.png",
      "https://images.hlido.eu/reviews/glips-figma-context-mcp/run-a626c0a0cef117fa-www-framelink-ai/pagedocs.png",
      "https://images.hlido.eu/reviews/glips-figma-context-mcp/run-a626c0a0cef117fa-www-framelink-ai/pagedocs_quickstart.png"
    ]
  },
  "pricing_facts": {
    "schema": "pricing-facts/1",
    "model": [
      "paid"
    ],
    "pricing_disclosed": {
      "pass": false,
      "evidence": "No pricing published; paid capability is a waitlist — \"Special features coming soon. Sign up for early access.\"",
      "tested_at": "2026-08-27"
    },
    "last_verified": "2026-08-27",
    "basis": "Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page.",
    "derived_at": "2026-08-27"
  }
}