社交媒体MCP服务器

social-mcp

An MCP (Model Context Protocol) server that lets AI assistants post and interact across social media platforms.

Supported Platforms

Quick Start

Run directly with npx — no install required:

npx social-mcp

Or install globally:

npm install -g social-mcp
social-mcp

MCP Client Configuration

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "social-mcp": {
      "command": "npx",
      "args": ["social-mcp"],
      "env": {
        "TELEGRAM_BOT_TOKEN": "your_token",
        "TWITTER_APP_KEY": "your_key",
        "TWITTER_APP_SECRET": "your_secret",
        "TWITTER_ACCESS_TOKEN": "your_token",
        "TWITTER_ACCESS_SECRET": "your_secret",
        "DISCORD_BOT_TOKEN": "your_token",
        "WHATSAPP_ACCESS_TOKEN": "your_token",
        "WHATSAPP_PHONE_NUMBER_ID": "your_id",
        "FACEBOOK_ACCESS_TOKEN": "your_token",
        "INSTAGRAM_ACCESS_TOKEN": "your_token",
        "SLACK_BOT_TOKEN": "your_token",
        "LINKEDIN_ACCESS_TOKEN": "your_token",
        "REDDIT_CLIENT_ID": "your_client_id",
        "REDDIT_CLIENT_SECRET": "your_client_secret",
        "REDDIT_USERNAME": "your_username",
        "REDDIT_PASSWORD": "your_password",
        "THREADS_ACCESS_TOKEN": "your_token",
        "THREADS_USER_ID": "your_user_id",
        "BLUESKY_IDENTIFIER": "your_handle",
        "BLUESKY_APP_PASSWORD": "your_app_password",
        "MASTODON_ACCESS_TOKEN": "your_token",
        "MASTODON_INSTANCE_URL": "https://mastodon.social",
        "TIKTOK_ACCESS_TOKEN": "your_token",
        "YOUTUBE_ACCESS_TOKEN": "your_token",
        "PINTEREST_ACCESS_TOKEN": "your_token",
        "MAIL_MAILER": "smtp",
        "MAIL_FROM_ADDRESS": "you@example.com",
        "MAIL_FROM_NAME": "Your Name",
        "MAIL_HOST": "smtp.gmail.com",
        "MAIL_PORT": "587",
        "MAIL_USERNAME": "you@gmail.com",
        "MAIL_PASSWORD": "your_password",
        "MAIL_ENCRYPTION": "tls",
        "SES_ACCESS_KEY_ID": "your_access_key_id",
        "SES_SECRET_ACCESS_KEY": "your_secret_access_key",
        "SES_REGION": "us-east-1",
        "DEVTO_API_KEY": "your_api_key",
        "HASHNODE_ACCESS_TOKEN": "your_token",
        "HASHNODE_PUBLICATION_ID": "your_publication_id",
        "BEEHIIV_API_KEY": "your_api_key",
        "BEEHIIV_PUBLICATION_ID": "your_publication_id",
        "GHOST_SITE_URL": "https://your-blog.ghost.io",
        "GHOST_ADMIN_API_KEY": "your_id:your_secret",
        "TWITCH_CLIENT_ID": "your_client_id",
        "TWITCH_CLIENT_SECRET": "your_client_secret",
        "TWITCH_ACCESS_TOKEN": "your_user_token",
        "TUMBLR_ACCESS_TOKEN": "your_oauth_token",
        "TUMBLR_BLOG_IDENTIFIER": "your-blog-name"
      }
    }
  }
}

Cursor / VS Code (via MCP extension)

{
  "mcp": {
    "servers": {
      "social-mcp": {
        "command": "npx",
        "args": ["social-mcp"],
        "env": { "...": "same env vars as above" }
      }
    }
  }
}

Only include env vars for the platforms you actually use — unconfigured platforms are reported at startup but do not prevent the server from running.

Environment Variables

Copy example.env to .env and fill in the credentials for the platforms you want to use.

Telegram

Twitter / X

Discord

WhatsApp Business

Facebook

Instagram

Slack

LinkedIn

Reddit

Threads

Bluesky

Mastodon

TikTok

YouTube

Pinterest

Medium

Dev.to

Hashnode

Beehiiv

Ghost

Twitch

Tumblr

Email

Email uses a Laravel-style driver system. Env vars are optional — credentials can also be passed inline per tool call (see Per-org / Multi-account Usage).

Set MAIL_MAILER to select your provider (smtp, sendgrid, mailgun, or ses) and only configure the variables for the driver you choose.

SMTP (MAIL_MAILER=smtp)

SMTP verification: On startup, social-mcp calls transporter.verify() against your SMTP server. A misconfigured host or wrong password shows [!!] in the startup log rather than failing silently on the first send.

SendGrid (MAIL_MAILER=sendgrid)

Mailgun (MAIL_MAILER=mailgun)

Amazon SES (MAIL_MAILER=ses)

Note: The sender address (MAIL_FROM_ADDRESS) must be a verified identity in your SES account. Sending uses the SES v2 REST API with AWS Signature V4 — no AWS SDK required. On startup (and via EmailService.verify()), your keys are checked with a signed GetAccount call, the same way SMTP is checked with transporter.verify(). To also check sandbox mode and 24h sending quota (e.g. before kicking off a bulk send), call EmailService.getSESAccountStatus() — it reuses that same signed request and returns { sandboxMode, max24HourSend, maxSendRate, sentLast24Hours }.

Available Tools

Telegram

• TELEGRAM_SEND_MESSAGE — Send a message to a chat or channel (chatId, text)
• TELEGRAM_SEND_PHOTO — Send a photo from a public URL or base64 bytes (chatId, photo, filename, caption)
• TELEGRAM_SEND_DOCUMENT — Send a file from a public URL or base64 bytes (chatId, document, filename, caption)
• TELEGRAM_GET_CHANNEL_INFO — Get channel metadata (channelId)
• TELEGRAM_FORWARD_MESSAGE — Forward a message between chats (fromChatId, toChatId, messageId)
• TELEGRAM_PIN_MESSAGE — Pin a message in a chat (chatId, messageId)
• TELEGRAM_GET_CHANNEL_MEMBERS — List channel administrators (channelId, limit)
• TELEGRAM_EDIT_MESSAGE — Edit the text of a message (chatId, messageId, text)
• TELEGRAM_DELETE_MESSAGE — Delete a message (chatId, messageId)

Note: Captions over 1024 characters (Telegram's limit) are automatically sent as a follow-up reply instead of failing the call.

Twitter / X

• SEND_TWEET — Post a tweet, optionally with up to 4 image attachments or 1 gif/video (text, max 280 chars, media)
• REPLY_TWEET — Reply to a tweet (tweetId, text)
• LIKE_TWEET — Like a tweet (tweetId)
• DELETE_TWEET — Delete a tweet (tweetId)
• GET_TWITTER_USER_INFO — Get a user's profile and metrics (username)
• SEARCH_TWEETS — Search recent tweets (query, maxResults)

Note: media takes { content, mimeType }[] with content as base64, uploaded via twitter-api-v2's uploadMedia (picks simple vs. chunked upload based on file size, so video is supported). TwitterCredentials also accepts a bearerToken in place of the four OAuth 1.0a fields (appKey/appSecret/accessToken/accessSecret) for callers that only have an OAuth 2.0 user-context token — X accepts it on both v2 and v1.1 media upload endpoints.

Discord

• SEND_DISCORD_MESSAGE — Send a message to a channel, optionally with file attachments (channelId, content, attachments)
• GET_DISCORD_MESSAGES — Retrieve recent messages from a channel, including attachment filenames/URLs (channelId, limit)

Note: attachments take { filename, content, contentType? } with content as base64 and are uploaded as real multipart attachments (not just an embedded image URL).

WhatsApp

• SEND_WHATSAPP_MESSAGE — Send a text, image, video, or document message (to, text, media, mediaKind, mediaContentType, mediaFilename)

Note: media is a public URL (sent as link) or base64-encoded bytes — base64 is uploaded to WhatsApp's media store first (/media) and referenced by the returned ID, since messages can't carry bytes inline. WhatsApp message retrieval requires webhook setup and cannot be polled via the API.

Facebook

• CREATE_FACEBOOK_POST — Create a post on a Facebook page, optionally with a photo (pageId, message, image, imageFilename)
• CREATE_FACEBOOK_VIDEO_POST — Upload and publish a video on a page (pageId, description, video, filename)
• GET_FACEBOOK_POSTS — Retrieve recent posts from a page (pageId, limit)

Note: image/video take either a public URL (fetched server-side via /photos' url / /videos' file_url) or base64-encoded bytes (uploaded as a real multipart source).

Instagram

• CREATE_INSTAGRAM_POST — Publish an image, video, or Reels post (userId, imageUrl, caption, mediaType: IMAGE/VIDEO/REELS)
• GET_INSTAGRAM_POSTS — Retrieve recent posts from an account, including media_url/permalink (userId, limit)

Note: imageUrl must be a publicly reachable URL for all media types — Instagram's Content Publishing API has no direct byte-upload path. For VIDEO/REELS, the container is polled until Meta finishes processing before publishing (previously publishing immediately could fail while the video was still processing).

Slack

• SEND_SLACK_MESSAGE — Send a message to a channel, optionally with file attachments (channelId, text, attachments)
• GET_SLACK_MESSAGES — Retrieve recent messages from a channel (channelId, limit)
• LIST_SLACK_CHANNELS — List public channels in the workspace (limit)

Note: attachments take { filename, content } with content as base64, uploaded via Slack's files.uploadV2. When present, text becomes the initial comment on the first file.

LinkedIn

• GET_LINKEDIN_PROFILE — Get the authenticated user's profile information
• CREATE_LINKEDIN_POST — Create a UGC post, optionally with an image (authorUrn, text, visibility, image, imageTitle)
• GET_LINKEDIN_POSTS — Retrieve recent posts by a member or organization (authorUrn, count)
• DELETE_LINKEDIN_POST — Delete a UGC post (ugcPostUrn)
• LIKE_LINKEDIN_POST — Like a post (actorUrn, ugcPostUrn)
• ADD_LINKEDIN_COMMENT — Comment on a post (actorUrn, ugcPostUrn, text)
• SEARCH_LINKEDIN_PEOPLE — Search for people by keywords (keywords, count)

Note: SEARCH_LINKEDIN_PEOPLE uses the LinkedIn People Search API which requires Partner Program access. Most developer apps will receive a 403. All other tools work with a standard OAuth 2.0 token. image is base64-encoded bytes — LinkedIn's UGC API has no URL-reference option, so it always goes through the register-upload → PUT bytes → reference-asset-URN flow.

Reddit

• REDDIT_SUBMIT_POST — Submit a text, link, or image post to a subreddit (subreddit, title, kind, text, url, image, imageFilename, imageMimeType)
• REDDIT_GET_POSTS — Get posts from a subreddit (subreddit, sort, limit)
• REDDIT_COMMENT — Comment on a post or reply to a comment (parentId, text)
• REDDIT_VOTE — Upvote, downvote, or remove a vote (id, direction)
• REDDIT_SEARCH — Search Reddit across all or a specific subreddit (query, subreddit, sort, limit)
• REDDIT_GET_USER_INFO — Get public info about a Reddit user (username)

Note: Reddit requires a "script" type app in reddit.com/prefs/apps. The password grant flow is used — 2FA must be disabled on the account. For kind: "image", base64 bytes go through Reddit's lease-then-S3-upload flow (/api/media/asset.json) and the post links to the resulting i.redd.it asset.

Threads

• THREADS_GET_PROFILE — Get your Threads profile information
• THREADS_CREATE_POST — Create a new post, optionally with an image or video (text, replyToId, mediaUrl, mediaType: IMAGE/VIDEO)
• THREADS_REPLY — Reply to a Threads post (replyToId, text)
• THREADS_GET_POSTS — Get recent posts from your account, including media_url (limit)
• THREADS_DELETE_POST — Delete a post by ID (mediaId)

Note: mediaUrl must be a publicly reachable URL, same as Instagram (shared Meta infrastructure). IMAGE/VIDEO containers are polled until processing finishes before publishing.

Bluesky

• BLUESKY_GET_PROFILE — Get your Bluesky profile information
• BLUESKY_CREATE_POST — Create a new post, optionally with up to 4 images (text, images)
• BLUESKY_REPLY_TO_POST — Reply to a Bluesky post (text, parentUri, parentCid, rootUri, rootCid)
• BLUESKY_GET_POSTS — Get recent posts from your feed (limit)
• BLUESKY_DELETE_POST — Delete a post by URI (uri, cid)
• BLUESKY_LIKE_POST — Like a post (uri, cid)
• BLUESKY_SEARCH_POSTS — Search for posts on Bluesky (query, limit)

Note: images take { content, mimeType, alt? }[] with content as base64 — AT Proto has no URL-reference option, every image is uploaded as a blob (uploadBlob) and referenced in the post's embed. For library consumers managing their own session (e.g. an app that already exchanged an identifier/app-password for a session and refreshes it independently), BlueskyCredentials also accepts { did, handle?, accessJwt, refreshJwt } to resume that session instead of logging in fresh — call getSession() afterward to read back the session in case AT Proto rotated the refresh token during the call, so you can persist the new one.

Mastodon

• MASTODON_GET_PROFILE — Get your Mastodon profile information
• MASTODON_CREATE_POST — Create a new post, optionally with up to 4 media attachments (status, inReplyToId optional, visibility, media)
• MASTODON_REPLY_TO_POST — Reply to a Mastodon post (status, inReplyToId)
• MASTODON_SEARCH_POSTS — Search for posts on your instance (query, limit)
• MASTODON_BOOST_POST — Reblog/boost a post (statusId)
• MASTODON_FAVOURITE_POST — Favorite/like a post (statusId)
• MASTODON_DELETE_POST — Delete a post (statusId)

Note: media take { content, filename?, description? }[] with content as base64, uploaded via /api/v2/media and attached by ID.

TikTok

• TIKTOK_QUERY_CREATOR_INFO — Query creator info including posting permissions and privacy options
• TIKTOK_GET_USER_INFO — Get TikTok account profile and stats
• TIKTOK_DIRECT_POST_VIDEO — Post a video by URL (videoUrl, title, privacyLevel, disableDuet, disableComment, disableStitch, brandOrganicToggle, brandContentToggle)
• TIKTOK_PHOTO_POST — Create a photo/carousel post (photoUrls, title, description, privacyLevel)
• TIKTOK_GET_POST_STATUS — Check the publishing status of a video or photo post (publishId)

Commercial content disclosure: brandOrganicToggle marks the post as promoting your own business (TikTok shows it as “Your brand”); brandContentToggle marks it as a paid partnership promoting someone else (“Branded content”). TikTok's API terms require one of these on any promotional post, and their app review checks that your posting UI offers the choice — so surface it to the user rather than hardcoding it. Both default to false, meaning not commercial content. TikTok rejects branded content on a private audience, so brandContentToggle with privacyLevel: "SELF_ONLY" is refused locally with an explanatory error instead of being sent and failing opaquely.

Note: TikTok's Content Posting API requires app approval from TikTok for Developers. Videos are pulled from a public URL by TikTok's servers (PULL_FROM_URL) rather than uploaded directly — TikTok's alternative FILE_UPLOAD mode (chunked byte upload for base64/local content) is not yet implemented here; if you need it, it's the same style of lease-then-PUT flow used for LinkedIn/Reddit media above.

YouTube

• YOUTUBE_GET_CHANNEL_INFO — Get channel info and stats (channelId optional — omit for your own channel)
• YOUTUBE_SEARCH_VIDEOS — Search YouTube for videos (query, maxResults, pageToken)
• YOUTUBE_GET_VIDEO_INFO — Get detailed info and stats for a video (videoId)
• YOUTUBE_LIST_CHANNEL_VIDEOS — List recent uploads from a channel (channelId, maxResults, pageToken)
• YOUTUBE_GET_COMMENTS — Get top-level comments on a video (videoId, maxResults)
• YOUTUBE_POST_COMMENT — Post a comment on a video (videoId, text)
• YOUTUBE_UPDATE_VIDEO — Update a video's title, description, and tags (videoId, title, description, tags, categoryId)
• YOUTUBE_UPLOAD_VIDEO — Upload a video to the authenticated user's channel (content, contentType, title, description, tags, categoryId, privacyStatus)

Note: YouTube requires an OAuth 2.0 access token with the https://www.googleapis.com/auth/youtube scope (or .upload for uploads). Enable the YouTube Data API v3 in your Google Cloud Console project. YOUTUBE_UPLOAD_VIDEO uses a single-request multipart upload (not YouTube's resumable/chunked protocol) — appropriate since content arrives as one base64 tool argument rather than a large local file; very large videos may need the resumable flow instead.

Pinterest

• PINTEREST_GET_BOARDS — List all boards on your Pinterest account (pageSize)
• PINTEREST_CREATE_BOARD — Create a new board (name, description)
• PINTEREST_CREATE_PIN — Create a pin on a board (boardId, title, description, image, imageContentType, link)
• PINTEREST_GET_PIN — Get details of a specific pin, including media URLs (pinId)
• PINTEREST_GET_BOARD_PINS — List all pins on a board (boardId, pageSize)
• PINTEREST_DELETE_PIN — Delete a pin by ID (pinId)

Note: image takes either a public URL or base64-encoded bytes — unlike most platforms, Pinterest's v5 API accepts base64 directly in the pin body (image_base64), no separate upload step needed.

Medium

• MEDIUM_GET_USER — Get your Medium profile and user ID (required for publishing)
• MEDIUM_CREATE_POST — Publish a Markdown article (authorId, title, content, tags up to 5, publishStatus, canonicalUrl)

Note: Medium Integration tokens are permanent and do not expire. Articles are published under your personal profile. To get your authorId, call MEDIUM_GET_USER first. The canonicalUrl field is useful for cross-posting from your own blog.

Email

• EMAIL_SEND — Send an email to a single recipient (to, subject, text, html, cc, bcc, replyTo, headers, attachments; plus optional inline credentials — see below)
• EMAIL_SEND_BULK — Send the same email to multiple recipients (recipients, subject, text, html, cc, bcc, replyTo, headers, attachments; plus optional inline credentials)

The active mail driver is selected by MAIL_MAILER. Switching providers requires only changing that one variable — no code changes needed. Credentials can also be passed inline per call for multi-account / per-org use cases. attachments take { filename, content, contentType? } with content as base64; for SES this transparently switches to a raw MIME message since attachments aren't supported by the plain send API.

Dev.to

• DEVTO_CREATE_ARTICLE — Create or draft a Dev.to article (title, bodyMarkdown, tags up to 4, published, description, canonicalUrl, series, mainImage)
• DEVTO_GET_MY_ARTICLES — List all articles published by the authenticated user (page, perPage)
• DEVTO_GET_ARTICLE — Get a specific article by ID including its full Markdown body (id)
• DEVTO_UPDATE_ARTICLE — Update an existing article (id, title, bodyMarkdown, published, tags, description, canonicalUrl)
• DEVTO_GET_ME — Get the authenticated user's profile — also useful to validate an API key at connect time

Note: Get your API key at dev.to/settings/extensions. Set published: false to save as a draft. mainImage is a URL only — Dev.to's public API has no image upload endpoint; body images must already be hosted and linked in the Markdown.

Hashnode

• HASHNODE_GET_PUBLICATION — Get publication info including ID, title, and URL
• HASHNODE_GET_POSTS — List recent posts from a Hashnode publication (first, publicationId)
• HASHNODE_CREATE_POST — Publish a post to a Hashnode publication (title, contentMarkdown, tags, subtitle, coverImageUrl, publicationId)

Note: Tags are passed as [{ name, slug }] objects. Get your publication ID via HASHNODE_GET_PUBLICATION or from the Hashnode dashboard URL. coverImageURL is a URL only — Hashnode's GraphQL API expects an already-hosted image, not a byte upload.

Beehiiv

• BEEHIIV_CREATE_POST — Create a newsletter post (title, bodyHtml, subtitle, status, audience, publicationId)
• BEEHIIV_GET_POSTS — List posts with send stats including open rate (page, limit, publicationId)
• BEEHIIV_GET_SUBSCRIBERS — List subscribers with status and tier info (page, limit, publicationId)

Note: Set status: "confirmed" to schedule the post for sending. status: "draft" saves it without sending. bodyHtml is raw HTML — Beehiiv's public API has no image/attachment upload; inline images must already be hosted.

Ghost

• GHOST_CREATE_POST — Create a post on a Ghost blog, optionally with a feature image (title, html, status, tags, excerpt, publishedAt, featureImage, featureImageFilename)
• GHOST_GET_POSTS — List posts filtered by status (page, limit, status)
• GHOST_UPDATE_POST — Update an existing post — requires the post's current updated_at timestamp for optimistic locking (id, updatedAt, title, html, status, tags, excerpt)
• GHOST_DELETE_POST — Permanently delete a post (id)

Note: The Admin API key (GHOST_ADMIN_API_KEY) must be in id:secret format from Ghost Admin → Settings → Integrations. Ghost uses JWT authentication which is generated automatically per request. featureImage takes a public URL or base64-encoded bytes; base64 is uploaded via Ghost's /images/upload/ endpoint first.

Twitch

• TWITCH_GET_USER — Get Twitch user info by username (login)
• TWITCH_GET_STREAMS — Get currently live streams, filtered by usernames or game ID (userLogins, gameId, first)
• TWITCH_GET_CHANNEL_INFO — Get channel details including current game and stream title (broadcasterId)
• TWITCH_SEARCH_CHANNELS — Search channels by name with optional live-only filter (query, liveOnly, first)
• TWITCH_SEND_CHAT_MESSAGE — Send a message to a Twitch chat channel (broadcasterId, senderId, message) — requires user OAuth token with chat:edit scope

Note: Most read tools use app access tokens generated automatically from TWITCH_CLIENT_ID + TWITCH_CLIENT_SECRET. TWITCH_SEND_CHAT_MESSAGE additionally requires TWITCH_ACCESS_TOKEN — a user-level OAuth token with the chat:edit scope.

Tumblr

• TUMBLR_GET_BLOG_INFO — Get blog info including title, description, post count, and follower count (blogIdentifier)
• TUMBLR_CREATE_POST — Create a post using the Neue Post Format, optionally with an image (text, title, tags, state, blogIdentifier, image, imageContentType)
• TUMBLR_GET_POSTS — Get posts from a blog, optionally filtered by type (offset, limit, type, blogIdentifier)
• TUMBLR_DELETE_POST — Delete a post by ID (postId, blogIdentifier)

Note: Tumblr uses OAuth 2.0 Bearer tokens. Set state: "draft" or "queue" to save without publishing immediately. image takes a public URL (referenced directly in the NPF content block) or base64-encoded bytes (uploaded inline via a multipart request with an identifier-referenced block).

Per-org / Multi-account Usage

Both email tools accept inline credentials as optional parameters. When provided, they take full priority over env vars — a fresh, isolated EmailService is created for that call. This enables per-organization or per-user email accounts without any env var changes.

Example — SMTP:

{
  "mailer": "smtp",
  "fromAddress": "team@acme.com",
  "fromName": "Acme Corp",
  "smtpHost": "smtp.gmail.com",
  "smtpPort": 587,
  "smtpUsername": "team@acme.com",
  "smtpPassword": "app-password",
  "smtpEncryption": "tls",
  "to": "customer@example.com",
  "subject": "Hello",
  "text": "Hi there!"
}

Example — SendGrid:

{
  "mailer": "sendgrid",
  "fromAddress": "team@acme.com",
  "sendgridApiKey": "SG.xxxx",
  "recipients": ["a@example.com", "b@example.com"],
  "subject": "Newsletter",
  "text": "Plain text body",
  "html": "<p>HTML body</p>"
}

Example — Mailgun:

{
  "mailer": "mailgun",
  "fromAddress": "team@acme.com",
  "mailgunApiKey": "key-xxxx",
  "mailgunDomain": "mg.acme.com",
  "to": "customer@example.com",
  "subject": "Hello",
  "text": "Hi there!"
}

When mailer and fromAddress are omitted, the tool falls back to the env var configuration (MAIL_MAILER, MAIL_FROM_ADDRESS, etc.).

Proxy Support

Most services accept an optional proxyUrl in their credentials (e.g. new DiscordService({ botToken, proxyUrl: "http://user:pass@host:port" })), routing that account's API calls through an HTTP(S) proxy — useful for a multi-tenant host that wants each connected account's traffic to originate from a distinct, consistent IP. This is a constructor-level option for programmatic/library consumers (not an MCP tool parameter, since it's server-operator infrastructure rather than message content).

Supported today: Discord, Slack, Telegram, Twitter/X, Instagram, Facebook, WhatsApp, Threads, LinkedIn, Reddit, Pinterest, Mastodon, YouTube, Ghost, Tumblr, TikTok, Medium, Bluesky. Not yet wired: Dev.to, Hashnode, Beehiiv, Twitch — these currently ignore any proxyUrl passed to them.

Three mechanisms are used depending on how each service talks to its platform, since passing the wrong one is a silent no-op rather than an error:

• createProxyDispatcher() (undici ProxyAgent) for services using raw fetch() — most of the list above.
• createProxyAgent() (https-proxy-agent) for services built on SDKs that take a classic http.Agent (Slack's @slack/web-api, Telegram's telegraf, Twitter's twitter-api-v2 via its httpAgent client option).
• Bluesky's BskyAgent takes a custom fetch function directly (AtpAgentOptions.fetch) — BlueskyService wraps createProxyDispatcher()'s dispatcher in one when proxyUrl is set.

createProxyDispatcher/createProxyAgent are exported from social-mcp/dist/lib/proxy.js if you need to build the agent/dispatcher yourself.

Development

pnpm install
pnpm run build      # compile TypeScript
pnpm run clean      # remove dist/
pnpm run watch      # watch mode
pnpm run test       # run tests
pnpm run lint       # lint with Biome
pnpm run format     # format with Biome

For a clean rebuild:

pnpm run clean && pnpm run build

Support

If you find this project useful, consider supporting its development:

License

MIT