MCP YouTube Transcript
An MCP server retrieving transcripts of YouTube videos
作者提供的说明(英文原文)
Schedule, cross-post, and analyze social media posts across 15 platforms from AI agents.
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。
BULKPUBLISH_API_KEY, BULKPUBLISH_TOOL_PROFILE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 …”。
适用于网页版、桌面版和手机版 Claude:
https://mcp.bulkpublish.com/mcp打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"bulkpublish": {
"command": "npx",
"args": [
"-y",
"@bulkpublish/mcp-server"
],
"env": {
"BULKPUBLISH_API_KEY": "your-value",
"BULKPUBLISH_TOOL_PROFILE": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add --transport http bulkpublish https://mcp.bulkpublish.com/mcpclaude mcp add bulkpublish --env BULKPUBLISH_API_KEY=your-value --env BULKPUBLISH_TOOL_PROFILE=your-value -- npx -y @bulkpublish/mcp-server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": {
"bulkpublish": {
"url": "https://mcp.bulkpublish.com/mcp"
}
}
}粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"bulkpublish": {
"command": "npx",
"args": [
"-y",
"@bulkpublish/mcp-server"
],
"env": {
"BULKPUBLISH_API_KEY": "your-value",
"BULKPUBLISH_TOOL_PROFILE": "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": {
"bulkpublish": {
"type": "http",
"url": "https://mcp.bulkpublish.com/mcp"
}
}
}粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"bulkpublish": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@bulkpublish/mcp-server"
],
"env": {
"BULKPUBLISH_API_KEY": "your-value",
"BULKPUBLISH_TOOL_PROFILE": "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。
ChatGPT 只能连接互联网上的服务器:
https://mcp.bulkpublish.com/mcp菜单名称可能因套餐和版本不同而略有差异。
你的选择通过 Cookie 保存在这台设备上,无需账户。
BULKPUBLISH_TOOL_PROFILE env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。
我们还没有读取这个服务器的工具列表;我们会轮流检查各个端点。
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7日
注册名称已通过 GitHub 账户 azeemkafridi 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
尚未检查;我们会轮流检查各个端点
尚未检查;我们会轮流检查各个端点
尚未检查;我们会在检查端点时读取工具列表
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 1 项适用的基础检查中通过了 1 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:BULKPUBLISH_API_KEY
两种方式都可以:安装在你的电脑上,或者连接云端版本。
GitHub 仓库最后更新于–,最新版本(1.42.0)于30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:MCP YouTube Transcript, Google Analytics MCP, BibiGPT YouTube Video Summarizer
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ 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
An MCP server retrieving transcripts of YouTube videos
GA4 MCP server for AI agents: schema discovery, server-side aggregation, analysis-ready defaults.
AI video summarizer: summarize and transcribe YouTube, Bilibili, TikTok and podcast videos.
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。
A Model Context Protocol server that lets Claude and other AI assistants interact with the BulkPublish social media publishing API.
This MCP server exposes BulkPublish API operations as tools that AI assistants can call directly. You can ask Claude to schedule posts, check analytics, upload media, manage labels, and more — all through natural conversation.
npx @bulkpublish/mcp-server
npm install -g @bulkpublish/mcp-server
bulkpublish-mcp
git clone https://github.com/azeemkafridi/bulkpublish-api.git
cd bulkpublish-api/mcp-server npm install npm run build node dist/index.js
Requires Node ≥ 20.19.
npm run buildbundles the composer UI with Vite (build:ui), then compiles the server withtsc.
Set your API key as an environment variable:
export BULKPUBLISH_API_KEY=bp_your_api_key_here
Get your API key at app.bulkpublish.com/developer.
Add this to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": { "bulkpublish": { "command": "npx", "args": ["-y", "@bulkpublish/mcp-server"], "env": { "BULKPUBLISH_API_KEY": "bp_your_api_key_here" } } } }
If you installed from source:
{
"mcpServers": { "bulkpublish": { "command": "node", "args": ["/absolute/path/to/bulkpublish-api/mcp-server/dist/index.js"], "env": { "BULKPUBLISH_API_KEY": "bp_your_api_key_here" } } } }
Add the MCP server to your Claude Code configuration:
claude mcp add bulkpublish -- npx -y @bulkpublish/mcp-server
Set the environment variable in your shell profile or .env file:
export BULKPUBLISH_API_KEY=bp_your_api_key_here
The npx/stdio setup above runs the server locally. Web clients that cannot spawn a local process — claude.ai custom connectors and Smithery's gateway — connect to a hosted HTTP endpoint instead. Alongside the stdio bin, the server ships a Streamable HTTP transport (dist/http.js, run with npm run start:http):
| Endpoint | Purpose |
|---|---|
POST /mcp | MCP Streamable HTTP (stateless) |
GET /.well-known/mcp/server-card.json | Static metadata so registries can skip scanning |
GET /health | Liveness check |
It is multi-tenant — the process holds no API key. Each caller supplies their own bp_… key per request via the connection URL (?key=bp_…), a Smithery ?config= blob, or an Authorization: Bearer bp_… / X-BulkPublish-Key header. initialize, tools/list, and resources/* need no key (so scans and tool discovery work); a tools/call without one returns 401.
A Dockerfile is included:
docker build -t bulkpublish-mcp . && docker run -p 8080:8080 bulkpublish-mcp
Once deployed, add it to claude.ai → Settings → Connectors → Add custom connector as https:// /mcp?key=bp_…, or publish the https:// /mcp URL to Smithery.
The hosted server at https://mcp.bulkpublish.com/mcp supports enterprise-managed authorization (the MCP io.modelcontextprotocol/enterprise-managed-authorization extension). An organization on Claude Team or Enterprise can connect every member to BulkPublish at once, with no sign-in screen and no pasted API key: the workspace owner trusts the organization's issuer (its Claude organization, or its identity provider) under API → MCP Auto-connect in BulkPublish, and members are matched to the workspace by their verified email address. In Claude, an organization Owner sets it up under Organization settings → Connectors → Set up auto-connect.
For client developers: the token endpoint accepts grant_type=urn:ietf:params:oauth:grant-type:jwt-bearer with an identity assertion, from a public client identified by a Client ID Metadata Document (an https client_id). Access tokens from this grant last up to an hour and carry no refresh token. Dynamic client registration and the authorization-code flow are unchanged.
The server ships three tool profiles, chosen with BULKPUBLISH_TOOL_PROFILE:
| Profile | Tools | Default for |
|---|---|---|
core | 26 — channels, posts, media, analytics, labels, approvals, post comments, channel sync settings, and the interactive panels | The hosted HTTP server (dist/http.js), which is what the Claude and ChatGPT directories list |
inbox | 32 — everything in core, plus the Inbox and comment-reply tools: list_conversations, get_conversation, reply_to_conversation, update_conversation, reply_to_comment, moderate_comment | Opt-in on a hosted deployment (BULKPUBLISH_TOOL_PROFILE=inbox). The core tools in it are identical to the core profile's |
full | Every tool (including the Inbox and comment tools; label management, hashtag groups, templates, calendar notes, review links, client-connect links, recurring schedules, channel sets, RSS feeds, chunked uploads, quota) | The local stdio server (npx @bulkpublish/mcp-server) |
Set BULKPUBLISH_TOOL_PROFILE=full on a hosted deployment to expose everything, or core locally to keep a small tool list. The panels' buttons call back into create_post, publish_post, retry_post, update_post, delete_post, list_channels, list_media, get_post, create_media_upload and finalize_media_upload, so all of those are in core.
Every tool carries a title and explicit readOnlyHint, destructiveHint, idempotentHint and openWorldHint booleans; npm run check:annotations verifies that for every profile (add --table to print the per-tool justification table used in directory submissions).
Tools marked ★ are in the core profile (see Tool profiles); tools marked ✉ are in inbox (with every ★ tool); the rest need full.
| Tool | Description |
|---|---|
| Posts | |
create_post ★ | Create and optionally schedule a post (supports reels, stories, carousels, threads via postTypeOverrides) |
compose_post ★ | Open an interactive composer UI (MCP Apps) to draft/schedule a post; submits via create_post |
update_post ★ | Update a draft or scheduled post: content, timing, media, labels, the full channel list (add or drop channels), per-platform text, thread parts, assignee and recurring schedule |
get_post ★ | Get a single post with full details |
list_posts ★ | List posts with filters (status, search, date range, labels, recurring) |
delete_post ★ | Delete a post |
publish_post ★ | Publish a draft post immediately |
retry_post ★ | Retry a failed post |
approve_post ★ | Approve a post awaiting team approval (roles with post:approve) |
reject_post ★ | Reject a pending post back to draft, with an optional reason |
get_post_metrics ★ | Get engagement metrics (impressions, likes, comments, shares). Each platform entry carries supportedMetrics — a key not in that list is a stored 0, not a measurement |
publish_story | Publish as a story to Facebook or Instagram |
bulk_posts | Bulk delete, retry or reschedule up to 100 posts |
get_queue_slot ★ | Get the next optimal time slot for a channel |
| Channels | |
list_channels ★ | List all connected social media channels |
get_channel_health | Check channel token health |
get_channel_options ★ | Get platform-specific options (boards, playlists) |
update_channel ★ | Turn an X channel's analytics sync or DM/reply reading on or off |
search_mentions | Search users for @mention (X, Bluesky, Mastodon, LinkedIn Pages — LinkedIn results carry the vanityName slug they matched on) |
| Channel Sets | |
list_channel_sets | List saved channel groups for one-click multi-channel targeting |
create_channel_set | Save a named group of channels (max 50 per org, names unique per org) |
update_channel_set | Rename a set or change its channels |
delete_channel_set | Delete a channel set |
| RSS Autopost | |
list_rss_feeds | List RSS/Atom feeds polled every 15 minutes (new items become posts) |
create_rss_feed | Add a feed (max 20 per org); mode = draft (default, items become drafts) or publish (auto-published); optional fieldMapping controls how items render (caption template, media pick, truncation, per-channel overrides) |
update_rss_feed | Change, pause, or re-point a feed (changing feedUrl re-baselines it — the backlog is not flooded) |
delete_rss_feed | Stop and remove a feed |
| Media | |
upload_media ★ | Upload a media file from a URL (or local path on the stdio server) |
get_media | Get a media file by ID |
list_media ★ | List uploaded media files |
delete_media | Delete a media file |
create_media_upload ★ | Reserve a presigned R2 URL for a direct browser upload (used by the composer) |
finalize_media_upload ★ | Record an uploaded object as a media file after the browser PUT (used by the composer) |
create_multipart_upload | Start a chunked upload for large media (videos up to 1GB) — presigned URLs for fixed 10MB parts |
complete_multipart_upload | Assemble the uploaded parts (partNumber + ETag each) and record the media file |
abort_multipart_upload | Cancel an in-progress chunked upload and free its stored parts |
| Labels | |
create_label | Create a new label |
list_labels ★ | List all labels |
update_label | Update a label name or color |
delete_label | Delete a label |
| Analytics | |
get_analytics ★ | Get analytics summary for a date range |
| Schedules | |
list_schedules | List recurring schedules |
create_schedule | Create a recurring schedule |
update_schedule | Update a schedule |
delete_schedule | Delete a schedule |
| Inbox and comments | |
list_conversations ✉ | List direct messages, reviews and comment threads, newest activity first, with filters and paging (Pro and Business plans) |
get_conversation ✉ | Read one conversation's messages, oldest first; does not mark it read |
reply_to_conversation ✉ | Send a reply to the person on the platform right away; it cannot be unsent |
update_conversation ✉ | Mark read or unread, archive or reopen, snooze, or assign a conversation; nothing is sent |
list_post_comments ★ | List the comments and reactions on a published post, per channel, with the ids the reply and moderate tools take |
reply_to_comment ✉ | Post a public reply to a comment right away, as the channel the post went out on |
In hosts that support MCP Apps — Claude, ChatGPT, VS Code, Goose, Postman, and the Smithery playground — the `compose_post` tool opens an interactive composer rendered inline in the conversation: pick channels, write content, optionally schedule, then submit. On
只显示了开头部分;其余内容请在仓库中阅读