Social Media MCP Server and Skill for Claude

What is going viral on TikTok in fitness right now, and which of it is still climbing?

Claude calls the radar for TikTok, filters by your keyword, and answers with the posts that gained the most in the last day, each with its growth and a link.

When should I post on Instagram this week?

The best hours per weekday, measured from the viral posts we hold for that network, with the hit rate and the worst hour to avoid.

Give me the 10 hashtags and 3 sounds that are working on Reels for coffee shops.

Hashtags ranked by the posts that used them and their engagement, plus the trending sounds, each with a link to hear it.

Which news stories are being shared the most today, and on which networks?

The articles from our feeds with the most Facebook, Reddit, Bluesky and Hacker News signal, so you write about what people are already talking about.

Turn the top post into a stat card I can publish.

Claude fetches one of our layout templates, fills it with the number and the source, and shows you the finished card right in the conversation.

Schedule that card to Instagram and X for Thursday at 6 pm.

On a paid plan with your accounts connected, Claude schedules or publishes it and reports the status per network.

Schedule one stoicism quote a day for Instagram and X at 9:00, public domain only, on the studio-quote-portrait template, with a two-line caption about what the author lived through.

The agent creates the recipe, then each day picks an unused quote, fills the template, renders the card and schedules the post.

Use my own template "Brand quote dark" and alternate it with studio-quote-photo; mix motivation and discipline.

Templates you customized carry your note for the agent; it follows it over any general rule and keeps your logo and colours as saved.

Every ten minutes, review what my team left in Review against the network rules, leave a note on each post, and give the OK that schedules it.

A reviewer agent of its own, with its own token. The content rule lives on the platform and the agent reads it on every pass: network terms, rights, facts with a DOI for any scientific claim, the picture against the text, spelling, and a risk grade from 1 to 10 on every post. It judges what each account will actually publish, the short copy for X or Bluesky included, not only the base text. Start it in verdict-only mode and let its OK publish once its judgement matches yours; the server still refuses an OK on a post the same user wrote, under 15 minutes away, or edited after the review. Team shows the flags per collaborator by week, fortnight and month.

Leave every post I ask for in Drafts, complete, and tell me in five lines where it is: id, project, link, angle, what you verified. I edit it and schedule it.

The working agreement teams settle on. Every post the agent makes answers with its project and a link that opens it from anywhere; the long notes stay on the post, for whoever reviews it, never in the chat.

Every morning, turn the top rising post in my niche into a photo-hook card and put it on the board for me to approve.

Same loop, different source: the radar instead of the quotes base. Nothing publishes until you approve the card.

My brand publishes in Spanish and in English. Make every post once, in Spanish, and when I approve it put the English version out too, image included.

One project per language, linked once in Projects. The agent works in one of them; on approval the other language comes out by itself: the copy, the first comment, the thread and the text inside the image are adapted (not word by word, hashtags in the new language, links and numbers kept), the same template re-renders with the translated words, and every network that cuts gets its own short copy. A picture with text baked in is the one thing it cannot translate, and it says so.