MCP (Model Context Protocol) lets Claude drive SocialCTL directly. Instead of switching to a web app, you ask Claude in plain language — "draft this week's calendar," "publish the approved carousel," "update the brand voice to bolder" — and Claude calls the SocialCTL MCP server to do it. This is what it means to manage social media from inside a chat.
What MCP is, briefly
MCP is an open standard for connecting AI assistants to external tools. A service exposes an MCP server that lists the actions it supports; the assistant can then call those actions with your permission. SocialCTL ships one, so Claude becomes a front-end to the whole content pipeline.
The mental model that helps: MCP is to AI assistants what an API is to code. An API lets one program call another; MCP lets a conversation call a program. Because it's a shared standard rather than a one-off integration, the same SocialCTL server works from any MCP-capable client — you're not locked into a single chat app. And because the assistant sees a described list of actions, it can chain them: draft a calendar, then generate the reels on it, then queue them, all inside one request.
What you can do from Claude
The SocialCTL MCP server exposes the core jobs of an AI social media manager:
- Compose & generate — draft calendar slots, generate a reel, carousel, or static post from a prompt.
- Publish & schedule — push approved content into the scheduling queue across the 8 supported networks.
- Brand kit — read and edit your brand kit: voice, colors, handles.
- Calendar — view, edit, re-roll, or regenerate your content calendar.
A sample chat session
You: Draft a 6-post calendar for my coffee shop this week, then generate the two reels.
Claude: (calls the calendar action) Drafted 6 slots — two reels, two carousels, two static. (calls generate on the reels) Both reels are rendering; I'll flag when they're ready.
You: Make the brand voice a bit bolder and regenerate the captions.
Claude: (edits the brand kit, re-rolls the captions) Done — voice set to Bold, captions refreshed.
Everything Claude did maps to a real SocialCTL action, charged the same credits and logged the same way as if you'd clicked it in the app.
Web app vs MCP, side by side
| Task | Web app | From Claude (MCP) |
|---|---|---|
| Draft a calendar | Fill the brief form | Ask in one sentence |
| Generate a reel | Click Generate on a slot | "Generate the reel for Tuesday" |
| Edit brand voice | Open brand-kit settings | "Make the voice bolder" |
| Publish | Approve, then schedule | "Schedule the approved posts" |
| Cost | Per-action credits | Identical per-action credits |
Why this matters
For anyone who already lives in Claude, MCP removes the context-switch. You plan, generate, and ship without leaving the conversation — the assistant becomes the operator, and SocialCTL is the hands. For agencies, it means a manager can drive a whole client roster conversationally, then let clients approve via no-login links.
There's a subtler benefit too: conversation is a better interface for iteration than forms. Refining content is naturally back-and-forth — "shorter," "more playful," "swap the second slide," "actually make all of them about the launch." In a web app each of those is a click-and-wait cycle; in a chat they're a sentence each, and the assistant holds the context of what you're building. The generation still runs through the same SocialCTL pipeline underneath, so you get the conversational front-end without giving up the brand kit, the calendar, or the credit accounting.
What MCP does not change
It's worth being clear about the boundaries. MCP is an interface, not a bypass. It doesn't skip approval — content still needs sign-off before it publishes. It doesn't skip credit costs — a reel generated from Claude is the same ~10 credits as one generated in the app. And it doesn't remove your control — Claude proposes and executes actions, but you're authorizing a server scoped to your own workspace, and every action lands in the same audited log. If anything, driving from a chat makes the audit trail clearer, because the conversation itself records what you asked for.
Setup
Connect the SocialCTL MCP server to your Claude client, authorize it against your workspace, and Claude can call the actions above. From there it's just chat. The authorization is scoped to your workspace, so an assistant can only ever act on your own content — the same isolation that keeps workspaces separate applies to MCP access.
Try SocialCTL first
Whether you drive it from Claude or the web app, the pipeline is the same. Generate a free sample calendar to see the content it produces, then wire up MCP when you want to run it from a chat.