# Social Expert: TikTok research and draft feedback

Publisher: ForUsApps, LLC. Independent service; not affiliated with TikTok.
Website: https://socialexpert.app/
Setup and supported goals: https://socialexpert.app/tiktok-mcp
Client kit: https://socialexpert.app/integrations
Capability JSON: https://socialexpert.app/mcp.json
Hosted Streamable HTTP: https://api.socialexpert.app/functions/v1/mcp
Official Registry: app.socialexpert/social-expert

## Fit the service to the task

- Regional TikTok format research: get_trending_formats, then get_trend_details using a returned ID. Associated sound details use get_sound_details; this is not an arbitrary music search.
- Posting windows: get_posting_times with the audience market and display timezone. Regional guidance, not private follower analytics.
- Keyword/hashtag research: get_keyword_insights and suggest_hashtags for the video's actual topic. Relative indicators, not exact search volumes.
- Content planning: get_content_ideas returns topic-based inspiration. A niche adaptation is editorial, not proof the niche performed well.
- Uploaded MP4/MOV draft feedback: open_video_analyzer, preview and explicit Analyze, then get_video_analysis for a retained report. No arbitrary public TikTok URL downloader, posting, scheduling or video editing.

A Social Expert account and user-authorized OAuth connection are required. There are 10 one-time free research lookups and 3 one-time free video reviews; paid monthly plans are available. Web and connected chat clients share balances. Reopening retained feedback uses no new review. Videos and reports are retained for 14 days. Never put account tokens, provider keys or backend secrets in a shared client configuration.

## Claude Code

Run this after reviewing the endpoint and requested access:

    claude mcp add --transport http social-expert https://api.socialexpert.app/functions/v1/mcp

Use /mcp to check the server and complete OAuth when prompted. For project configuration, merge the social-expert entry from claude-code.json into .mcp.json; do not overwrite existing servers. Official client reference: https://code.claude.com/docs/en/mcp

## VS Code

Run MCP: Add Server, select HTTP and enter the hosted endpoint. Alternatively merge the servers entry from vscode.json into the appropriate MCP configuration. Follow the client's normal trust and OAuth flow. Official reference: https://code.visualstudio.com/docs/agents/reference/mcp-configuration

These examples follow current client configuration documentation; this kit does not claim a fresh end-to-end test of every client/version. Company policies and client plans can limit custom connectors. ChatGPT and Claude public directory submissions are still under review; custom connection does not imply approval.

## First useful request

"Research three TikTok format ideas for a US home bakery. I can film kneading dough and decorating one cake. Separate returned research from your creative suggestions. Give me one premise, first frame and shot list per idea. Ask before starting video analysis."

Useful context: subject, audience market, display timezone when requesting times, and filming constraints. Returned source timestamps may be unknown; don't infer that an example is currently growing. Exact source URLs and dated observations are preferable to reach promises.

## Evidence and workflow examples

- Original dated example with underlying public research and TikTok links: https://socialexpert.app/learn/still-frame-signals-tiktok-trend
- Public research snapshot: https://socialexpert.app/research/2026-10-30-still-frame-signals.json
- Planning recipe: https://socialexpert.app/workflows/tiktok-content-plan
- Draft review recipe: https://socialexpert.app/workflows/tiktok-draft-review
- Privacy: https://socialexpert.app/privacy.html
- Terms: https://socialexpert.app/terms.html
- Support: support@forusapps.com