What the SocialGPT MCP server does
The SocialGPT MCP server gives an AI agent access to social media data held in SocialGPT. It can inspect analyzed videos, retrieve detailed video analysis, search content, and produce a content profile covering topics, pillars, and voice. Account tools expose connected profiles, performance metrics, follower history, growth summaries, and the changing metrics of an individual post.
The SocialGPT MCP server also supports public creator research. An agent can retrieve a creator by platform and handle, list that creator’s recent or top-performing videos, request analysis where available, and find videos similar to a specified video. Public or competitor analysis does not require a connected social account, while analysis of your own content requires a SocialGPT account with connected sources.
How it works
SocialGPT is provided as a hosted remote MCP endpoint rather than a local package. Clients connect over Streamable HTTP using:
https://mcp.gpt.social/mcp
Authentication uses OAuth 2.1 with PKCE. The default access model is read-only analytics. Separate scopes control ingesting new content for analysis and publishing to connected accounts. The whoami and server_info tools can identify the authenticated caller and report server scope information.
Some operations are asynchronous. For example, analyze_creator returns a job_id; the agent can pass that identifier to get_analysis_status to check progress. Publishing also provides get_publish_status for polling a publish job.
Setup and configuration
Register the endpoint in an MCP client and complete the SocialGPT OAuth sign-in flow when prompted. The README provides configurations for Claude Desktop, Claude Code, Cursor, VS Code, ChatGPT, Codex, Cline, and Windsurf. Other MCP-compatible clients can use the same URL with Streamable HTTP support.
No local installation or API key is required according to the project documentation. A SocialGPT account is needed for authenticated access, and connected social accounts unlock analysis of owned content. Publishing requires the optional content:publish scope. Before publishing to TikTok, call get_publish_options; it returns current privacy and interaction settings and indicates whether posting is available.
Tools and capabilities
The SocialGPT MCP server exposes tools for:
- Listing, searching, retrieving, and deeply analyzing videos.
- Reading connected accounts and account-level performance history for up to 365 days.
- Tracking follower, following, and media counts over time.
- Calculating growth changes and momentum for a selected window.
- Looking up public creators, their videos, and similar videos.
- Analyzing a TikTok, YouTube, or Instagram post from its URL.
- Starting creator analysis jobs and checking their status.
- Uploading local video files through an upload page, listing uploaded drafts, and passing a draft ID to publishing.
- Publishing to connected TikTok, Instagram, and YouTube accounts.
Publishing behavior varies by platform. TikTok supports drafts or direct publishing, while Instagram publishes reels and carousels immediately and YouTube publishes videos immediately.
Limitations and notes
Read access, content ingestion, and publishing are governed by different OAuth scopes. Publishing tools are the only tools that act on connected accounts, and the documentation states that explicit confirmation is required before anything goes live. Metric history is limited to a trailing window of up to 365 days.
The server does not describe local execution, environment-variable configuration, or a self-hosted deployment path. Public creator analysis is available without connecting an account, but analyzing your own content depends on connected social accounts and the relevant access scope. The SocialGPT MCP server is therefore best suited to agents that can use remote MCP connections and complete browser-based OAuth authentication.