SocialGPT MCP Is Here: Your Social Analytics Inside Claude, ChatGPT, and Any AI Agent

Today we're launching the SocialGPT MCP server — a direct line between your social media data and the AI tools you already use every day. Connect it once, and Claude, ChatGPT, Cursor, or any MCP-compatible agent can pull your analytics, video breakdowns, and competitor research into the middle of a conversation. No exports, no copy-paste, no switching tabs.

Ask "why did my last reel underperform?" and your AI agent answers with your actual numbers, your actual transcript, and a comparison against videos like yours — not generic advice.

What is MCP, in one paragraph

The Model Context Protocol is an open standard that lets AI assistants connect to external tools and data sources. Think of it as a universal port: instead of every app building a custom integration for every AI assistant, a service exposes one MCP server and every compatible client — Claude, ChatGPT, Cursor, Codex, VS Code, and a fast-growing list of others — can plug in. SocialGPT's server lives at https://mcp.gpt.social/mcp and works with all of them.

What your AI agent can do once connected

The server exposes 15+ tools across four areas, and your agent picks the right ones automatically based on what you ask:

• Your content — list and search your analyzed posts, pull deep video analysis (hooks, scene-by-scene breakdowns, transcripts, themes), and track views, engagement, and follower growth over any window across TikTok, Instagram, and YouTube.
• Competitor intel — look up any public creator's profile and content DNA, browse their top performers, find videos similar to yours, and benchmark your numbers against comparable content. New creators and posts can be analyzed on demand.
• Your content profile — a synthesized picture of your content pillars, voice, and recurring themes that evolves as you post, so the agent's suggestions actually sound like you.
• Guided workflows — pre-built prompts that chain tools together: a full account audit, a head-to-head benchmark against similar videos, or a reverse-engineered playbook for any creator's niche. In Claude these show up right in the composer under the SocialGPT connector.

How to connect it

Everything below uses the same server URL: https://mcp.gpt.social/mcp. Pick your client, connect, and sign in with your SocialGPT account when the browser window opens. If you don't have an account yet, create one free and connect a social account first so there's data to analyze.

Claude (claude.ai)

1. Open Settings → Connectors → Add custom connector.
2. Name it SocialGPT and paste the server URL.
3. Click Connect and approve access with your SocialGPT login. The tools — and the guided workflow prompts — are now available in any conversation via the + menu.

Claude Desktop

Add this to your config file (on macOS: ~/Library/Application Support/Claude/claude_desktop_config.json), then restart the app:

{
  "mcpServers": {
    "socialgpt": {
      "url": "https://mcp.gpt.social/mcp"
    }
  }
}

Claude Code

One command in your terminal:

claude mcp add --transport streamable-http socialgpt https://mcp.gpt.social/mcp

Bonus: the guided workflows become real slash commands in Claude Code — type /mcp__socialgpt__analyze_my_account and hit enter.

ChatGPT, Cursor, and everything else

ChatGPT accepts the server URL as a connector in developer mode, and Cursor takes a three-line entry in .cursor/mcp.json. Codex, AntiGravity, VS Code, and other MCP clients follow the same pattern: point them at the URL, authenticate, done. Copy-paste configs for every client are on our MCP integration page.

Five things to try first

• "How did my content perform this month compared to last?"
• "Analyze my latest reel and tell me why it did or didn't work."
• "Find creators in my niche on TikTok and show me what's working for them."
• "Compare my best video against similar videos and tell me how to beat them."
• "Based on my top performers, write five hooks for my next post."

That last one is where this gets fun: the agent grounds its creative suggestions in your real performance data — the hooks that already worked for you, the formats your audience finishes, the topics that consistently land.

Security: read-only, OAuth, revocable

Connecting an AI agent to your data should not require trust falls. The SocialGPT MCP server authenticates with OAuth 2.1 with PKCE — you approve access through the same login you already use, and no API keys ever get copied into config files. Every scope is read-only today: a connected agent can analyze your content but cannot post, edit, or delete anything. And you can revoke any connection at any time from your SocialGPT settings.

What's next

Analysis is step one. On the roadmap: write capabilities so your agent can draft, schedule, and publish through the same connection, plus companion skills that pair SocialGPT's insights with design tools like Canva — so "turn my best-performing post into a carousel for Instagram" becomes a single sentence instead of an afternoon.

Connect it today: grab the URL, pick your client on the setup page, and ask your AI agent something only your data can answer.