Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者による説明(英語の原文)
Ask your AI to check, fix and submit your App Store release. 41 tools for the paperwork.
この3つを1つのスコアにまとめることはしません。意見、測定、人気はそれぞれ別のものだからです。
このサーバーのやさしい解説はまだ書いていません。以下はレジストリから読み取った事実です。
使っているAIアプリを選んで、設定をコピーしてください。
ASC_KEY_ID, ASC_ISSUER_ID, ASC_PRIVATE_KEY_PATH, ASC_LICENSE_KEY「秘密」と表示された値は、あなたのアカウントの鍵です。絶対に人に教えたり、チャットに貼り付けたりしないでください。
必要なもの
どのアプリを使っていますか?
上でアプリを選んでください。選んだアプリを覚えておくので、次からはボタンに「…にインストール」と表示されます。
Claude Desktop → Settings → Developer → Edit Config を開き、claude_desktop_config.json にこれを貼り付けます
{
"mcpServers": {
"asc": {
"command": "npx",
"args": [
"-y",
"@pofky/asc-mcp"
],
"env": {
"ASC_KEY_ID": "your-value",
"ASC_ISSUER_ID": "your-value",
"ASC_PRIVATE_KEY_PATH": "your-value",
"ASC_LICENSE_KEY": "your-value"
}
}
}
}ターミナルでこれを実行します
claude mcp add asc --env ASC_KEY_ID=your-value --env ASC_ISSUER_ID=your-value --env ASC_PRIVATE_KEY_PATH=your-value --env ASC_LICENSE_KEY=your-value -- npx -y @pofky/asc-mcp動かす前に、空欄の値を入力してください。
プロジェクトの .cursor/mcp.json、または ~/.cursor/mcp.json に貼り付けます
{
"mcpServers": {
"asc": {
"command": "npx",
"args": [
"-y",
"@pofky/asc-mcp"
],
"env": {
"ASC_KEY_ID": "your-value",
"ASC_ISSUER_ID": "your-value",
"ASC_PRIVATE_KEY_PATH": "your-value",
"ASC_LICENSE_KEY": "your-value"
}
}
}
}ブラウザがアプリを開いてよいか確認し、そのあとアプリが設定を表示して確認を求めます。キーの値は空欄のままなので、自分で入力してください。
プロジェクトの .vscode/mcp.json に貼り付けます
{
"servers": {
"asc": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@pofky/asc-mcp"
],
"env": {
"ASC_KEY_ID": "your-value",
"ASC_ISSUER_ID": "your-value",
"ASC_PRIVATE_KEY_PATH": "your-value",
"ASC_LICENSE_KEY": "your-value"
}
}
}
}ブラウザがアプリを開いてよいか確認し、そのあとアプリが設定を表示して確認を求めます。キーの値は空欄のままなので、自分で入力してください。
これはパソコンで動くサーバーですが、ChatGPTがつなげるのはインターネット上のサーバーだけです。ほかのアプリを選んでください。
選んだアプリは、この端末のCookieに保存されます。アカウントは不要です。
この設定は、作者のレジストリ登録情報から作成しています。READMEと違う場合は、READMEに従ってください。
パソコンで動くサーバーです。インストールすると、ツール一覧がAIアプリに表示されます。
誰でも見られる情報をチェックしており、コードは読んでいません。当てはまらない項目は数に入れていません。 · 2026年11月9日にチェック
レジストリ名は、リポジトリを所有するGitHubアカウント pofky で確認済みです
このリポジトリはまだGitHubから読み取っていません
このリポジトリはまだGitHubから読み取っていません
これはパソコンで動くため、インターネット上のエンドポイントはありません
パソコンで動き、代わりにローカルの設定を使います
パソコンで動くサーバーです。ツール一覧を読み取れるのは、インターネット上のエンドポイントがある場合だけです
レジストリでこのカテゴリにある有効なすべてのサーバーと比較しています。スターは、公開者がリポジトリの所有者である場合だけ数えます。
当てはまる基本チェック1項目のうち1項目に合格しました。コードは読んでいません。キーやトークンを追加する前に、サーバーがどんなアクセスを求めているかを確認し、必要な権限だけを持つキーを使ってください。
入力が必要なもの:ASC_KEY_ID, ASC_ISSUER_ID, ASC_PRIVATE_KEY_PATH, ASC_LICENSE_KEY
あなたのパソコンで動きます。必要なときにAIアプリが起動し、データは作者のサーバーを通りません(サーバー自体がオンラインサービスを呼び出す場合を除きます)。
GitHubリポジトリの最終更新は–、最新バージョン(1.10.3)がレジストリに届いたのは5 日前です。
Claude Desktop、Claude Code、Cursor、VS CodeなどのMCP対応アプリで使えます。パソコンにNode.jsも必要です。
同じカテゴリには:Spec Driven Development, Remote Desktop Commander, Mcphost
同じカテゴリで、ほかの作者が作ったサーバー。スターの多い順。
カードの見方:↓ 週あたりのダウンロードi★ GitHubスターi3つの柱:レビュー · チェック · 利用状況i更新=最新バージョンをレジストリに公開した日i
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Hosted MCP runtime where your agent publishes its own tools by tool call and keeps working.
作者のリポジトリから書かれたとおりにコピーしたものです。PopNx MCPは内容を確認・推奨していません。HTML、画像、スクリプトは取り除いています。リンクはこのサイトの外へ移動します。
Ship an App Store release from your coding agent. An MCP server for App Store Connect: 41 tools that edit version metadata, upload screenshots, attach builds, drive TestFlight, create in-app purchases, submit and release, plus an intelligence layer that audits a version before you submit and summarizes your reviews. 6 slash-command workflows and a bundled Claude Skill. Not another API wrapper.
asc-mcp.pages.dev · npm · Pricing: 7-day free trial, then $9/month, VAT included (sales tax added in the US, Canada and India)
You made a toy and you want it in a giant toy store, the App Store. The store has a strict office. Before your toy reaches the shelf, you fill out lots of forms, add pictures, set a price, and wait for a helper to check it all. Forget one thing and they say "no, try again", and days are gone.
asc-mcp is a helper robot for that office. It talks to your AI assistant, so you just say what you want:
Checking whether the store said yes or no is free, with no time limit. Everything else works free for 7 days, no card, then it's $9 a month, VAT included, with sales tax added in the US, Canada and India. The robot checks your homework, and the store still decides.
For the grown-ups: the toy is your app, the office is App Store Connect, and the robot is an MCP server running on your own computer.
Claude Desktop, one click: download the latest asc-mcp.mcpb and open it. Claude asks for your Issuer ID and lets you pick your .p8 with a file picker. No JSON, no restart hunt.
Everything else (Claude Code, Cursor, Windsurf, Cline), one command:
npx @pofky/asc-mcp init --write # finds your .p8, asks for your Issuer ID, writes your MCP config
Then ask your agent: "start my asc-mcp trial, my email is you@example.com". All 41 tools unlock for 7 days, no card and nothing to cancel, and the trial activates in the session you are already in, so whatever you were blocked on works on the very next call. After that Pro is $9 a month (VAT included; in the US, Canada and India sales tax is added at checkout). Six tools stay free, and two of those (the setup check and the playbook) need nothing but Node. The Pro tools that only read (reviews, sales, builds, TestFlight groups, preflight, briefing, keywords, competitors, metadata diff, release notes) also work 5 times in total on your machine with no license, as a preview.
| You say | What happens |
|---|---|
| "Run a preflight check on my app" | Audits metadata, character limits, screenshots and build status, and says plainly when a check could not run. |
| "Give me a morning briefing" | Summarizes all your apps: who's in review, who got rejected, new low-rating reviews, action items. |
| "Generate release notes from my git history" | Reads commits since last tag, categorizes them, and gives you structured data to write "What's New" text. |
| "Is my app in review?" | Exact review state with context ("typical time: 24-48 hours") |
| "Show me 1-star reviews" | Customer reviews filtered by rating, territory, sorted by date |
| "What were my downloads this week?" | Sales and revenue summary by territory |
No context switching. No portal. Just ask.
Maintained successor to JoshuaRileyDev/app-store-connect-mcp-server (archived Feb 2026). Different angle, same API surface plus more.
Other ASC MCP servers wrap the API and give you 80 to 982 raw endpoints. This one gives you 41 opinionated tools, 6 slash-command Prompts, and a Claude Skill that all think. The read tools summarize and audit; the Pro control tools actually drive App Store Connect, from editing metadata to submitting and releasing, with the same API key. Two tools use MCP Sampling: your own client's model does the LLM work, so there is no extra cost from this server.
Why full control is possible without fastlane: fastlane's deliver and pilot are just calls to the same App Store Connect REST API this server authenticates against with your .p8. So nearly everything you would script with fastlane is a tool here, no Ruby toolchain required. The only step that needs your Mac is building and signing the binary, which build_and_archive and upload_binary drive via Xcode.
You need: an Apple Developer account, Node 18 or newer, and an AI coding app that supports MCP (Claude Code, Claude Desktop, Cursor, Windsurf, Cline or any other MCP client). MCP is the standard way these apps plug in extra tools; asc-mcp is one of those tools, and it runs on your own computer.
Step 1. Create an App Store Connect API key. Open App Store Connect > Integrations > App Store Connect API and generate a key with the Admin or App Manager role. From that same page you need two things:
AuthKey_XXXXXXXXXX.p8, contains the Key ID.Move the key into Apple's standard folder, where asc-mcp finds it on its own:
mkdir -p ~/.appstoreconnect/private_keys
mv ~/Downloads/AuthKey_*.p8 ~/.appstoreconnect/private_keys/
Step 2. Run init.
npx @pofky/asc-mcp init --write
It finds the key, asks for your Issuer ID, backs up your Claude Desktop / Claude Code config and adds the server to it, so there is no JSON to edit. Leave out --write to print the block and paste it yourself.
Step 3. Restart your agent and check. Ask it: "run asc_setup_check". It checks the key, the Issuer ID, a live connection to App Store Connect and your license, and prints the exact fix for anything wrong. The same check from a terminal is npx @pofky/asc-mcp doctor. Then ask: "List my App Store Connect apps".
Step 4. Start the trial. Ask your agent: "start my asc-mcp trial" and give it your email. It calls asc_start_trial, which unlocks all 41 tools for 7 days with no card and nothing to cancel. The key works in the running session, so whatever you were blocked on works on the very next call, and it is saved on your machine, in ~/.asc-mcp/license.json and in any MCP client config that already has an asc-mcp entry, so it survives a restart. One trial per Apple developer account. Without a trial, six tools stay free with no time limit, and most Pro tools that only read work 5 times in total as a preview.
Claude Desktop, no terminal. Download the .mcpb bundle from the latest release and open it. Claude asks for your Issuer ID and lets you pick the .p8 with a file picker.
As a Claude Code plugin. One line installs the server and the review-triage skill together, and Claude Code asks for your Issuer ID when it enables the plugin:
/plugin install asc-mcp --marketplace pofky/asc-mcp
The plugin lives in plugin/, and its README says what it runs and sends. In your agent, the /asc-start command walks a first-time user through setup.
Running init from an agent, or any pipe. There is no terminal to answer questions in, so pass what would have been asked:
npx @pofky/asc-mcp init --write --issuer <your-issuer-uuid> # add --config <path> if you keep several client configs
It never invents a value: if the Issuer ID is missing it says so instead of writing a config that fails at Apple's auth. --key-path, --key-id and --license are there for the same reason, and every one of them also reads from the matching ASC_* environment variable. Already paid? --license adds your Pro key to the config.
Global install. npm install -g @pofky/asc-mcp, then use asc-mcp where the examples say npx -y @pofky/asc-mcp.
Manual config. Add to your agent's MCP config (Claude Desktop config, or ~/.claude.json for Claude Code):
{
"mcpServers": { "appstore-connect": { "command": "npx", "args": ["-y", "@pofky/asc-mcp"], "env": { "ASC_ISSUER_ID": "YOUR_ISSUER_ID" } } } }
With the .p8 in ~/.appstoreconnect/private_keys/, ASC_ISSUER_ID is the only required env var (Key ID and key path are auto-detected). Add ASC_LICENSE_KEY to unlock Pro. A key kept elsewhere: set ASC_PRIVATE_KEY_PATH to the file and ASC_KEY_ID to its 10-character ID.
Works with Claude Code, Cursor, Windsurf, Cline, and any MCP-compatible client.
| Tool | What it does |
|---|---|
asc_start_trial | 7 days of everything, no card. Ask your agent to run it with your email. Unlocks all 41 tools in the running session, no restart, nothing to cancel. |
asc_setup_check | Run first if anything's off. Checks your key, Issuer ID, a live App Store Connect connection, and license tier, and prints the exact fix for each failure. |
asc_guide | Start here. The in-agent playbook for every flow (first release, update, IAP, subscriptions, reviews, TestFlight, binary), with each manual App Store Connect website step flagged inline. See also USER_GUIDE.md and LIMITATIONS.md. |
list_apps | List all your apps with name, bundle ID, SKU and App ID |
app_details | Version history, build status, release state, dates |
review_status | Current review state with human-readable context |
$9 per month, VAT included; in the US, Canada and India sales tax is added at checkout, before you pay. Cancel any time.
| Tool | What it does | Why it matters |
|---|---|---|
list_reviews | Customer reviews filtered by rating, territory, sort order | See what users say without opening the portal |
sales_report | Daily/weekly/monthly downloads and revenue by territory | Know your numbers instantly |
release_preflight | Pre-submission audit: metadata, char limits, screenshots, builds | Finds submission blockers before you press Submit, and names any check it could not run |
daily_briefing | Morning summary across all apps: status, reviews, rejections | One call replaces 10 minutes of portal clicking |
release_notes | Git commits since last tag, categorized for writing "What's New" | Your AI agent writes release notes from your actual changes |
keyword_insights | Analyze keywords against iTunes search competition, difficulty ratings | See which keywords are worth targeting at a glance |
competitor_snapshot | Look up any app: ratings, reviews, version, price, release notes | Competitive intelligence without leaving your editor |
metadata_diff | Compare live vs pending version metadata across all locales | Verify exactly what changed before submitting |
triage_reviews | Pulls recent reviews and clusters them into 3 to 5 themes with counts, action buckets, and quotes, using MCP Sampling | Zero extra cost: your client's model does the clustering |
| draft_review_response | Drafts a public reply to a single review in the review's locale, via Sampling. Elicits tone if your client supports it. Never auto-posts. | Apple gu
冒頭だけを表示しています。続きはリポジトリで読めます