Ask your AI agent how your UGC performed
Weekly UGC reporting usually means exporting numbers from four platforms and lining them up by hand. With UGC API's MCP server you ask your agent instead: "how did we do this week?", "which caption won?", "why did posts fail?". It answers from one views number across TikTok, Instagram, Facebook and YouTube, with a table, a chart and a step you approve.
Updated
Connect once
- Add
https://ugcapi.dev/mcp/v2as a custom connector in Claude or as an MCP server in ChatGPT, Cursor, VS Code or Gemini CLI. The directory listing at/mcpworks too, without the summary tools and results card. - Sign in with your UGC API account when the OAuth window opens. The agent sees the accounts and posts of that workspace only.
- Ask a question. Read-only tools (reports, summaries, failures) don't spend credits.
Questions and the tools behind them
| You ask | Tool | You get |
|---|---|---|
| How did our UGC do this week? | get_results_summary | Published, views, best platform by views per post, top clips, breakouts, spend |
| Which platform won the serum launch? | get_campaign_report | Platforms ranked by average views, lift over the median, engagement, cost per 1,000 views, the winner |
| Which caption or hook worked? | get_campaign_report with group_by: variant | Caption variants side by side, with how far ahead the winner is |
| Which creator should get more clips? | get_campaign_report with group_by: account | Accounts ranked with engagement and breakouts |
| How did this clip do everywhere? | get_clip_performance | One clip on every platform and account, views at 1 h to 7 d, links |
| Why did posts fail? | explain_failures | Errors grouped, the reason in plain words, the fix, and what can be retried |
What the answer looks like
- A one-line answer first, for example "TikTok won: 2,400 views per post, 2.3× Instagram."
- A table of platforms, clips, captions or failures, from the Markdown summary each tool returns.
- A results card in apps that support MCP Apps: KPI tiles and a bar chart of what won, drawn from the same data.
- A next step to approve: repost the breakout clip to backup creators, switch next week's main caption, or retry what failed. Nothing posts until you say yes.
Why the numbers line up
Every clip keeps one clip_id on every account and platform, and views are one number: TikTok views, Reel plays and Shorts views. When a platform doesn't report something, the value is null and the agent says "not reported" instead of showing a zero. See campaign reports.
FAQ
Does asking for reports cost credits?
No. Only posting spends credits. Reports, summaries, clip lookups and failure explanations are free.
Which agents can show the results card?
Apps that support MCP Apps or ChatGPT apps render it inline. Everywhere else the agent shows the Markdown table and summary from the same result.
Can the agent act on the results?
Yes, after you approve: it can repost a clip, schedule next week's campaign with the winning caption, or retry failed posts, using the same credits as the API.