Automate Twitter/X Posts and Searches

Twitter/X Automation via Rube MCP automates Twitter/X operations through Composio's Twitter toolkit. It requires Rube MCP connected (https://rube.app/mcp, no API keys needed) with an active Twitter connection via RUBE_MANAGE_CONNECTIONS (OAuth), and RUBE_SEARCH_TOOLS must be called first for current tool schemas.

Use it across six workflows: creating/deleting/looking up posts (TWITTER_CREATION_OF_A_POST, limited to 280 weighted characters where some Unicode counts as more than one and URLs count as a fixed 23 characters), searching posts (TWITTER_RECENT_SEARCH for the last 7 days only, TWITTER_FULL_ARCHIVE_SEARCH requiring Academic Research or Enterprise access for older tweets, and TWITTER_RECENT_SEARCH_COUNTS to get a match count without fetching the tweets themselves - empty results return meta.result_count: 0 with no data field rather than an error), looking up users by username/ID/batch, managing bookmarks (private, requiring the authenticated user's numeric ID), managing lists (owned/membership/pinned/followed, also keyed by numeric user ID), and liking/unliking posts. Media attachment follows a fixed sequence:

1. Upload media with TWITTER_UPLOAD_MEDIA (images) or TWITTER_UPLOAD_LARGE_MEDIA (video/GIF)
2. Get media_id from response
3. Pass media_id as string in media__media_ids array to TWITTER_CREATION_OF_A_POST

Search queries use Twitter's own operator syntax: from:username, to:username, @username mentions, #hashtag, quoted exact phrases, has:media/has:links, is:retweet/-is:retweet, is:reply/-is:reply, and lang:en, combined with space for AND, OR for either, - prefix for NOT, and parentheses for grouping. Nearly every workflow starts with TWITTER_USER_LOOKUP_ME to get the authenticated user's numeric ID, since most endpoints require it rather than a username - usernames themselves must be passed without the @ prefix and are case-insensitive. User and media IDs are numeric strings, not integers, throughout the API.

The most critical gotcha: post creation is not idempotent - retrying on a timeout creates duplicate posts, so automated posting needs its own deduplication logic. Rate limits vary significantly by access tier (Free tier allows as few as 1,500 posts/month), checkable via the x-rate-limit-remaining response header. Batch user lookup (TWITTER_USER_LOOKUP_BY_IDS) accepts a maximum of 100 IDs per request, and suspended or deleted accounts return errors rather than empty results. Bookmark pagination uses pagination_token while search pagination uses next_token - a naming inconsistency worth tracking when building generic pagination handling across endpoints.

Teams automating Twitter/X posting, search, and engagement through an MCP-connected agent who need the character-limit, idempotency, and ID-format gotchas already documented to avoid duplicate posts or silent lookup failures. A full quick-reference table maps every task directly to its tool slug and required parameter - e.g. TWITTER_POST_DELETE_BY_POST_ID takes id, TWITTER_ADD_POST_TO_BOOKMARKS takes tweet_id, TWITTER_LIST_LOOKUP_BY_LIST_ID takes list_id - so the right call can be made without re-deriving the correct workflow sequence each time.