Automate Social Media Posting Across Platforms

Publora Skills is a set of nine official SKILL.md files that teach Claude Code, Codex, Cursor, OpenClaw, or any skills-aware agent how to publish and schedule posts on LinkedIn, X, Instagram, Threads, TikTok, YouTube, Facebook, Bluesky, Mastodon, and Telegram through the Publora MCP server. The pitch: your agent drafts, you approve, Publora publishes - one integration instead of ten platform APIs, each with its own auth, media pipeline, and rate limits. The MCP server itself exposes 16 tools.

Use it when an agent needs to post or schedule content across multiple social platforms without you wiring up ten separate platform APIs, or when it needs to react to, comment on, or reshare LinkedIn posts. The free Starter plan covers every platform except X, whose API costs are billed per call and passed through as a paid-plan requirement.
Know each skill's specific limits before relying on it: linkedin-post can't build a swipeable carousel (the LinkedIn API reserves those for sponsored content - use a PDF document for swipeable multi-page content instead); instagram-post needs a JPEG and a Business account, since Instagram's Graph API refuses PNG and text-only posts; tiktok-post will publish as private (SELF_ONLY) for any app that hasn't passed TikTok's audit, regardless of the requested visibility; threads-post can't auto-split long content into a connected thread, since Threads currently disables that; and telegram-post is capped at a 1,024-character media caption and a 50 MB bot video ceiling.

The MCP server's 16 tools split into posts and media (list_connections, list_posts, create_post, get_post, update_post, delete_post, get_upload_url, complete_media, delete_media, prune_media_reference) and LinkedIn engagement (linkedin_create_reaction, linkedin_delete_reaction, linkedin_create_comment, linkedin_delete_comment, linkedin_create_reshare, linkedin_list_mentionables). create_post accepts up to 10 public https mediaUrls, which Publora downloads server-side and attaches before validation, so media and scheduling can go in a single call instead of a four-step upload flow. Omitting scheduledTime creates a draft; a future ISO 8601 UTC time schedules the post, and a time five or more minutes in the past is rejected. Platform IDs aren't guessable - call list_connections and copy the platformId values verbatim. update_post can patch content, platforms, scheduledTime, and platformSettings on a draft or scheduled post, so a typo doesn't require deleting and recreating it. Every MCP tool also has a REST twin at https://api.publora.com/api/v1, authenticated with an x-publora-key header instead of a bearer token - LinkedIn statistics and the per-platform platformSettings block (Instagram video type, TikTok privacy, YouTube visibility, Telegram delivery flags, Threads reply control) are REST-only:

curl -X GET "https://api.publora.com/api/v1/platform-connections" \
  -H "x-publora-key: sk_your_api_key"

Get an API key at app.publora.com/dashboard/api, then connect an agent. For any skills-aware agent:

npx skills add publora/skills

Claude Code can instead connect directly with claude mcp add publora --transport http https://mcp.publora.com --header "Authorization: Bearer sk_YOUR_KEY", or install the plugin via /plugin marketplace add publora/skills. Codex adds publora/skills as a marketplace and installs the publora-skills plugin; Cursor adds https://mcp.publora.com under Settings, MCP; Claude Desktop needs a config-file snippet on macOS or Windows; OpenClaw and the claude.ai connector (OAuth, no key needed) have their own setup docs. Update any time by re-running npx skills add publora/skills. The project is MIT licensed.

Anyone running an AI agent that needs to draft, schedule, and publish social content across multiple platforms with human approval in the loop, without maintaining ten separate platform integrations.