spotify-mcp
Enables AI assistants to search Spotify and manage playlists, library, and playback via a remote MCP server running on Cloudflare Workers.
README
spotify-mcp
A remote MCP (Model Context Protocol) server for Spotify, running on Cloudflare Workers. It lets an AI assistant search Spotify and manage your playlists, library and playback — including moving tracks around inside playlists.
Auth is handled entirely over the web: the server is its own OAuth provider to
the MCP client, and performs the Spotify OAuth flow upstream. There is no local
binary and no authenticate tool — connecting the server in your MCP client
opens the Spotify consent screen in your browser.
Architecture
McpAgent(SpotifyMCP, a Durable Object) hosts the MCP server and tools, served over Streamable HTTP at/mcp(and legacy SSE at/sse).@cloudflare/workers-oauth-providerwraps the Worker. It issues tokens to MCP clients and stores the upstream Spotify tokens (access + refresh) encrypted in the grantprops.SpotifyHandler(Hono app) implements/authorizeand/callback, driving the Spotify authorization-code flow and showing the consent dialog.- The agent persists a working access token in its Durable Object state and refreshes it automatically using the stored refresh token.
MCP client ──/mcp──▶ OAuthProvider ──▶ SpotifyMCP (Durable Object)
│ │ │
/authorize /callback Spotify Web API
└────── Spotify consent (browser) ────────┘
Tools
| Tool | What it does |
|---|---|
get_me |
Current user profile |
search |
Search tracks / albums / artists / playlists |
list_playlists |
Your playlists |
get_playlist_tracks |
Playlist details + tracks with their positions |
create_playlist |
Create a playlist |
update_playlist_details |
Rename / re-describe / change public state |
add_tracks_to_playlist |
Add tracks, optionally at a specific position |
remove_tracks_from_playlist |
Remove tracks |
reorder_playlist_tracks |
Move a track (or range of tracks) to a new position |
unfollow_playlist |
Unfollow / delete a playlist |
get_saved_tracks / save_tracks / remove_saved_tracks |
Liked-songs library |
get_playback_state |
What's playing, on which device |
control_playback |
play / pause / next / previous (Premium + active device) |
add_to_queue |
Queue a track |
Deploy it yourself
Runs on the Cloudflare Workers free tier.
Prerequisites
- A Cloudflare account and the Wrangler CLI logged in (
npx wrangler login). - A Spotify Developer account.
- Node.js 18+.
0. Get the code
git clone https://github.com/lassejlv/spotify-mcp.git
cd spotify-mcp
npm install --ignore-scripts # --ignore-scripts avoids an unused native build (sharp)
1. Create a Spotify app
At https://developer.spotify.com/dashboard, create an app and note the Client ID and Client Secret. Under Redirect URIs, add your Worker's callback URL:
https://spotify-mcp.<your-subdomain>.workers.dev/callback
(The workers.dev subdomain is shown after the first wrangler deploy. Add the
URI, then deploy again if needed. A custom domain's /callback works too.)
2. Create the KV namespace
The OAuth provider stores grants, tokens and clients in KV.
npx wrangler kv namespace create OAUTH_KV
Copy the returned id into wrangler.jsonc, replacing the existing OAUTH_KV
id value (the checked-in one belongs to the original author's account and won't
work for you).
3. Set secrets
npx wrangler secret put SPOTIFY_CLIENT_ID # your Spotify Client ID
npx wrangler secret put SPOTIFY_CLIENT_SECRET # your Spotify Client Secret
npx wrangler secret put COOKIE_ENCRYPTION_KEY # any random string, e.g. `openssl rand -hex 32`
Optional — restrict who can use the server:
# Comma-separated allowlist of Spotify account emails. Only these accounts can
# authorize; everyone else is rejected at the callback (no token issued).
# Leave unset to allow any Spotify account.
npx wrangler secret put ALLOWED_EMAILS # e.g. me@example.com,friend@example.com
4. Deploy
npx wrangler deploy
Your server is now live at https://spotify-mcp.<your-subdomain>.workers.dev/mcp.
Connect an MCP client
Point any remote-MCP-capable client at the /mcp URL. Clients that only speak
stdio can bridge via mcp-remote:
{
"mcpServers": {
"spotify": {
"command": "npx",
"args": ["mcp-remote", "https://spotify-mcp.<your-subdomain>.workers.dev/mcp"]
}
}
}
On first connect, your browser opens: approve the MCP client, then log in and grant Spotify access. Tokens are refreshed automatically thereafter.
Local development
cp .dev.vars.example .dev.vars # then fill in the three values
npx wrangler dev
wrangler dev simulates KV locally. Use http://localhost:8788/callback as an
additional Spotify redirect URI for local testing.
Notes
- Tracks and playlists can be referenced by bare ID or full
spotify:URI in any tool. reorder_playlist_tracksuses zero-based positions; callget_playlist_tracksfirst to see current positions.- Playback control endpoints require Spotify Premium and an active device.
- Requested scopes: playlist read/modify (public + private), library read/modify,
playback read/modify,
user-read-private, anduser-read-email(used for theALLOWED_EMAILSaccess gate). - Access control: set the
ALLOWED_EMAILSsecret to a comma-separated list to restrict the server to specific Spotify accounts; leave it unset to allow anyone.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。