Figma Context MCP
Give your coding agent access to your Figma data. Implement designs in any framework in one-shot.
作者提供的说明(英文原文)
SEO audit (400 checks), text readability and grammar check, PDF to images, image conversion.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
Authorization标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
这个地址需要在请求的 header 中放入密钥,而添加连接器的菜单无法设置。请改用 Claude Code、Cursor 或 VS Code。
在终端中运行这条命令
claude mcp add --transport http app-kit-dev https://mcp.app-kit.dev/mcp --header 'Authorization: your-value'运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"app-kit-dev": {
"url": "https://mcp.app-kit.dev/mcp",
"headers": {
"Authorization": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"app-kit-dev": {
"type": "http",
"url": "https://mcp.app-kit.dev/mcp",
"headers": {
"Authorization": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个地址需要在请求的 header 中放入密钥,而添加连接器的菜单无法设置。请改用 Claude Code、Cursor 或 VS Code。
你的选择通过 Cookie 保存在这台设备上,无需账户。
Authorization 机密 HTTP header这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
工具名称和描述与服务器在 2026年11月8日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。
text_checkCheck a text: metrics (filler words, spam, keyword stuffing), readability, stop phrases, grammar. Input: text up to 20000 characters, lang (language of the text) ru or en. Returns passed (blocking thresholds), metrics by layer, verdicts (blocking=false means a recommendation), to
seo_audit_fastQuick SEO audit of a public website (up to 10 pages): score 0-100 by check group, failed checks, top 5 losses. The site root is audited: URL path and query are dropped (https://example.com/page?x=1 -> https://example.com/). Waits up to 60 s for the result, sending progress. If no
seo_audit_statusStatus and result of an SEO audit by the handle from seo_audit_fast. Does not consume quota. While status=running, call again after poll_after_s seconds. A handle is valid for 60 min and only for the caller who started the audit.
pdf_to_imagesConvert an existing PDF to images (one file per page): format png/jpeg/webp, dpi 72-300, first_page/last_page. PDF up to 20 MB and up to 20 pages. Input: exactly one of source_url (public URL, the server downloads it), upload_id (create_upload, then PUT) or data_base64 (up to 512
images_to_pdfCombine existing images into one PDF (one image per page), page_size A4/Letter or the image size. Input: source_urls[] or upload_ids[], up to 20 files, each up to 20 MB. Output: download links (resource_link, valid for 10 min; deleted once all files of the job are downloaded) and
image_convertConvert an existing image to another format: target_format png/jpeg/webp/bmp/tiff/gif, quality 1-100, max_dim (downscale the long side). File up to 20 MB. For avif and maximum compression use image_compress. Input: exactly one of source_url (public URL, the server downloads it),
conversion_statusStatus and result of a conversion by the handle from pdf_to_images / images_to_pdf / image_convert. Does not consume quota. While status=running, call again after poll_after_s seconds. A handle is valid for 1 hour and only for the caller who started the conversion.
image_compressCompress an existing image to WebP or AVIF: quality 1-100, width/height (downscale), preset (photo/picture/drawing/icon/text), lossless. Returns a result link (valid up to 5 min), original and new size, compression ratio. File up to 20 MB. Input: exactly one of source_url (public
create_uploadCreate a one-time upload link for a local file (for clients with local file access, e.g. Claude Code, Cursor). Returns upload_id, put_url and a curl command: upload the file with a single PUT (up to 20 MB, within 10 min), then pass upload_id to a file tool. One upload_id serves o
quota_statusShow the current access tier and the remaining calls for each tool. Tiers: L0 anonymous (shared per platform for Claude and ChatGPT), L1 personal token, L2 App-Kit account (OAuth). Does not consume quota.
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月8日
注册名称已通过 GitHub 账户 app-kit-dev 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
在 613 毫秒内响应(HTTP 200)
无需登录
已读取 10 个工具描述,未发现针对 AI 的隐藏指令
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 3 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:Authorization
它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。
GitHub 仓库最后更新于–,最新版本(0.1.3)于34天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。
同一分类中有:Figma Context MCP, Claude Real Video, Figma MCP Server
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Give your coding agent access to your Figma data. Implement designs in any framework in one-shot.
Let any LLM watch a video locally — and search everything it has ever watched.
The Figma MCP server brings Figma design context directly into your AI workflow.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Remote MCP server for SEO audits, text readability and grammar checks, image conversion (PNG/JPEG/WebP) and PDF to images - for Claude, Cursor, VS Code, Windsurf and any other MCP client. Free, with daily limits. Two addresses:
https://mcp.app-kit.dev/mcp # no account: anonymous or with a free token
https://mcp.app-kit.dev/mcp/account # with a free App-Kit account (OAuth sign-in), higher limits
Русская версия | Tools reference | app-kit.dev
This repository holds the public documentation, client configs and registry metadata. The server itself is a hosted service operated by App-Kit; there is nothing to install.
There are two ways to connect:
https://mcp.app-kit.dev/mcp. Works anonymously out of the box; a free token (no email) gives you limits of your own.https://mcp.app-kit.dev/mcp/account. Accepts only account sign-in (OAuth 2.1 with PKCE): your client opens the App-Kit sign-in page in the browser, you sign in or create an account (email and password) and approve access. Highest limits, no tokens to copy. Create an account in advance at admin.app-kit.dev/oauth/signup or from the sign-in page your client opens.App-Kit, Remote MCP server URL: https://mcp.app-kit.dev/mcp/account. Click Connect and sign in. Details: examples/claude-desktop/claude.ai.md. claude mcp add --transport http app-kit https://mcp.app-kit.dev/mcp/account
Then run /mcp in Claude Code and choose Authenticate for app-kit.
Works anonymously out of the box. For higher limits, get a free token on app-kit.dev/mcp (one click, no email) and pass it either as a header Authorization: Bearer mcpf_... or in the URL: https://mcp.app-kit.dev/mcp/t/mcpf_....
claude mcp add --transport http app-kit https://mcp.app-kit.dev/mcp
claude mcp add --transport http app-kit https://mcp.app-kit.dev/mcp \ --header "Authorization: Bearer mcpf_your_token"
Add --scope user to make it available in every project. See examples/claude-code.sh.
Customize > Connectors > + > Add custom connector. Name: App-Kit, Remote MCP server URL: https://mcp.app-kit.dev/mcp/t/mcpf_your_token (the token in the URL is recommended: requests from claude.ai share Anthropic's IP addresses, so anonymous limits there are a small shared pool). Authentication: no sign-in. Free Claude plans allow one custom connector. Details: examples/claude-desktop/claude.ai.md.
.cursor/mcp.json in the project or ~/.cursor/mcp.json globally (example):
{
"mcpServers": { "app-kit": { "url": "https://mcp.app-kit.dev/mcp", "headers": { "Authorization": "Bearer mcpf_your_token" } } } }
.vscode/mcp.json (example); VS Code asks for the token once and stores it securely:
{
"inputs": [ { "type": "promptString", "id": "app-kit-token", "description": "App-Kit MCP free token (mcpf_...)", "password": true } ], "servers": { "app-kit": { "type": "http", "url": "https://mcp.app-kit.dev/mcp", "headers": { "Authorization": "Bearer ${input:app-kit-token}" } } } }
Open the MCP config from the Cascade MCP settings (mcp_config.json) and add a server with serverUrl. ${env:VAR} is expanded in headers:
{
"mcpServers": { "app-kit": { "serverUrl": "https://mcp.app-kit.dev/mcp", "headers": { "Authorization": "Bearer ${env:APPKIT_MCP_TOKEN}" } } } }
cline_mcp_settings.json (MCP Servers > Configure). Set "type": "streamableHttp": without it Cline uses the legacy SSE transport. Agent-oriented steps: llms-install.md.
{
"mcpServers": { "app-kit": { "type": "streamableHttp", "url": "https://mcp.app-kit.dev/mcp", "headers": { "Authorization": "Bearer mcpf_your_token" } } } }
Install as an extension (the repo root has gemini-extension.json):
gemini extensions install https://github.com/app-kit-dev/mcp
or add the server to ~/.gemini/settings.json (httpUrl = Streamable HTTP):
{
"mcpServers": { "app-kit": { "httpUrl": "https://mcp.app-kit.dev/mcp", "headers": { "Authorization": "Bearer mcpf_your_token" } } } }
#### Other clients
Any client that supports remote MCP over Streamable HTTP: URL https://mcp.app-kit.dev/mcp, optional header Authorization: Bearer mcpf_.... Clients that cannot set headers can use https://mcp.app-kit.dev/mcp/t/mcpf_.... Clients with MCP OAuth support can use https://mcp.app-kit.dev/mcp/account instead.
Full descriptions and JSON Schemas: docs/tools.md (generated from the live tools/list).
| Tool | What it does | Example prompt |
|---|---|---|
seo_audit_fast | Fast SEO audit of a site root (up to 10 pages): score, failed checks, top-5 losses. Waits up to 60 s, then returns a handle | "Run an SEO audit of example.com and tell me what to fix first" |
seo_audit_status | Status and result of an audit by handle | (called by the model automatically) |
text_check | Readability, water/spam metrics, stop phrases, grammar; lang ru/en, up to 20 000 chars | "Check this landing page copy for readability and grammar: ..." |
quota_status | Your access level and remaining free calls per tool | "How many free App-Kit checks do I have left today?" |
image_convert | Convert between PNG, JPEG, WebP, BMP, TIFF, GIF | "Convert this logo to PNG: https://example.com/logo.webp" |
pdf_to_images | Render PDF pages (up to 20) to PNG/JPEG/WebP | "Turn pages 1-3 of https://example.com/deck.pdf into PNG at 150 dpi" |
images_to_pdf | Merge up to 20 images into one PDF | "Make a PDF from these 3 screenshots: ..." |
conversion_status | Status and result of a long conversion by handle | (called by the model automatically) |
create_upload | One-time upload URL for a local file (agents with a shell) | "Convert ./scan.pdf to PNG pages" |
image_compress | Compress to WebP/AVIF with quality and resize options | "Compress https://example.com/hero.png to AVIF" |
All tools are read-only: nothing is published or changed on your side. Results that are files come back as short-lived download links (10 minutes) plus a small preview.
Limits are per tool, per day, and reset at 00:00 UTC.
| Tool | Anonymous | Free token | App-Kit account | Notes |
|---|---|---|---|---|
text_check | 3 / day | 5 / day | 20 / day | |
seo_audit_fast | 1 / day | 2 / day | 5 / day | one domain: up to 5 audits per day from all users |
seo_audit_status, conversion_status, quota_status | free | free | free | do not spend quota |
image_convert | 2 / day | 3 / day | 10 / day | |
pdf_to_images, images_to_pdf | 1 / day | 2 / day | 5 / day | up to 20 pages / files, 20 MB |
create_upload | 3 / day | 5 / day | 15 / day | up to 20 MB, link valid 10 min |
Anonymous requests from claude.ai share one small hourly pool (Anthropic's shared IP range). Every response includes quota (remaining, reset_at, tier, upgrade_url). There is also a shared daily budget per tool; when it runs out, the tool says so and resets at 00:00 UTC. Limits may be tuned without notice; the current values are always in the tool descriptions and quota_status.
The free tier above stays free. Higher limits, full multi-page SEO audits with PDF reports and larger files are part of the paid App-Kit API and are billed in App-Kit credits - see app-kit.dev.
text_check is not stored; only metrics are returned./mcp/account): email, password hash, sign-up date and source, stored in Germany; OAuth tokens are stored as hashes (access 1 hour, refresh 30 days). To delete your account, write to hi@wm2.dev.Full policy: app-kit.dev/privacy.
Do I need an account? No. Anonymous use works; a free token (no email) raises the limits. An App-Kit account (/mcp/account) gives the highest limits and signs you in through the browser.
Why do anonymous calls from claude.ai run out quickly? claude.ai connects from Anthropic's shared IP range, so all anonymous claude.ai users share one small pool. Use a token in the URL.
Can it audit my staging site? Only publicly reachable sites. Private networks, localhost and non-standard ports are blocked.
Which MCP protocol versions are supported? Streamable HTTP, stateless: current MCP revisions and clients on 2025-11-25. No stdio package is needed.
A long audit returned `status: running`. What now? The model calls seo_audit_status with the handle after poll_after_s seconds. The handle works only for the caller and expires in 60 minutes.
Is the server open source? The server is a hosted service. This repository (docs, examples, registry metadata) is MIT-licensed.
Where do I get help? Email hi@wm2.dev (questions, account deletion, problems with the server). Bugs and ideas: issues. Security: see SECURITY.md.
只显示了开头部分;其余内容请在仓库中阅读