DesignIntel

It can write great code — it just can't actually see your design. So you keep paying for it in time, money, and rework.

- close A generic Figma MCP server hits Figma's API on every prompt — burn through the rate limit mid-session and your AI just stalls.
- close Figma's own Dev Mode MCP needs a paid Dev Mode seat, and only sees whatever file happens to be open in the desktop app right now.
- close Nothing set up yet? Your AI is guessing paddings and colors from a screenshot, and you're the one fixing it by hand.

- check Sync once, then every query comes from your own local cache — no rate limits, no re-fetching, no waiting mid-session.
- check No desktop app to keep open, no Dev Mode seat required — just a personal access token, working the same whether Figma's running or not.
- check Nothing is uploaded, ever. Your designs stay cached on your machine, not ours.

One command to install, thendesignintelto start it — no account, no signup, no desktop install. Paste a Figma file link and it copies to your machine: pages, components, styles and variables, indexed for instant search.

Paste one line into Cursor, Claude, or Copilot. That's the whole setup — no accounts to provision, no seats to buy for the team.

Your AI reads the real design as often as it needs — instant, private, and with zero request limits. What ships finally matches the mockup.

This is what actually happens once you're connected — no screenshots, no guessing, no rate limits.

Using the designintel MCP server, sync figma.com/design/aBcDeFg123/Marketing-Site, then build the "Pricing" section as a React + Tailwind component. Pull the exact colors, spacing, and typography from the synced file — don't guess or approximate.

→ sync_file("aBcDeFg123") ✓ already cached

→ search("Pricing section") ✓ found frame 42:108

→ get_node("42:108") ✓ 14 nodes, resolved styles

What comes back — real values, not guesses

export function Pricing() { return ( <section className="bg-[#0B0B0F] px-6 py-24"> <div className="grid gap-8 md:grid-cols-3"> <div className="rounded-2xl border border-white/10 p-8"> <p className="text-sm font-semibold text-[#818CF8]">Starter</p> <p className="mt-4 text-4xl font-bold">$0</p> </div> </div> </section> ); }

No cloud middleman. No third-party storage. No seat you have to buy for everyone on the team. DesignIntel runs entirely on your machine — which is exactly why it can be unlimited.

computer Local-first all_inclusive Unlimited requests payments No per-seat cost

A production-quality local cache with full-text search — for the developers who want the details.

Everything is served from a local cache, so your AI can query designs as often as it wants — no rate limits, no throttling.

Your design data lives on your own disk. It never touches a third-party server.

Add every Figma account you use — client work, your own, a team's — right from the dashboard. Still no cloud, still local-only.

Search across every cached node, component and style instantly — no API round-trip.

Compact, simplified responses designed to minimize the tokens each AI tool call burns.

Pull shared styles and design variables so your AI writes code that matches your system.

Needs Node.js 20+. Just want to try it without installing anything?npx designintelworks too — same thing, one-off. Works on macOS, Windows and Linux.

No token needed yet — open the dashboard, add a Figma personal access token right there, and you're set. Add more than one if you work across multiple teams or clients.

3. Add the endpoint to your AI tool's config

{ "mcpServers": { "designintel": { "url": "http://localhost:8787/mcp" } } }

The dashboard athttp://localhost:8787shows this same config, plus a paste-a-link sync flow and account management if you'd rather not touch a terminal again after this.

DesignIntel is a local MCP server that syncs your Figma files to your own machine and serves them to AI coding tools like Cursor, Claude, and Copilot — so your AI builds from your real design (components, spacing, colors and variables) instead of guessing from a screenshot. One command to install, no account required.

Any AI coding tool that speaks MCP — including Cursor, Claude, GitHub Copilot, VS Code, Windsurf, Zed and Cline. You add one line to your tool's config and you're done.

Yes — it's free. Install it withnpxand run it with your own Figma token; there's nothing to sign up for.

No. There's no login and no account. DesignIntel reads a Figma personal access token from an environment variable and runs entirely on your machine.

Completely. DesignIntel runs entirely on your machine. Your Figma files are stored on your own disk and are never uploaded to a third-party server.

Do I need a paid seat for every developer? add

No. DesignIntel doesn't charge per seat and doesn't cap how many requests your AI can make — everything is served from a local copy, so it stays unlimited and instant for your whole team.

Does DesignIntel need to keep running? add

Yes — the dashboard and MCP endpoint only respond while it's running, so install and run it once first before opening the dashboard. Turn on auto-start from the dashboard's Settings page (or rundesignintel enable-autostart) and it launches automatically every time you log in, so you never have to remember to start it again.

Local-first, private by design, and free — no account required.

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

Extentos is a multi-vendor development platform for adding smart-glasses capabilities to existing iOS and Android apps. The simplest analogy is Stripe for smart glasses

An MCP server tailored for React Native–first development using Gluestack UI

Create and read feature flags, review experiments, generate flag types, search docs, and interact with GrowthBook's feature flagging and experimentation platform.

Gives AI agents public URLs (tunnels) for localhost, live HTTP traffic inspection, snapshot publishing, and access control.

Understand, develop, and debug authorization policies in Oso Cloud.

Up-to-date documentation for your coding agent. Covers 1000s of public repos and sites. Built by ref.tools

Points agents to the canonical Agentry install docs for errors, product analytics, and deploy attribution.

Gives coding agents deterministic, read-only documentation handoffs with edit boundaries, ownership, and required checks before repository changes.

Deterministic JSON checks with signed evidence and x402-paid execution

Search Apple's Developer Documentation with smart search and wildcard support.