CreatorScore MCP Server
Vet creators, influencers, and athletes from inside Claude Desktop, Claude Code, ChatGPT, or any MCP client — without leaving the conversation.
This is a Model Context Protocol server. It's a thin client over the public CreatorScore API. It holds exactly one secret — your API key — and makes plain HTTPS calls to https://creatorscore.io/api/v1/*.
It has no database access and no scraping/ML/provider credentials. Every score is produced server-side by CreatorScore's gated v2.2 seven-agent scoring pipeline — the same one the brand dashboard uses. The MCP user is strictly less privileged than a logged-in brand: credit metering, rate limits, the daily fresh-scrape circuit breaker, and tenant isolation all apply automatically.
Install
- npm:
npx -y creatorscore-mcp (set CREATORSCORE_API_KEY)
- Claude Desktop / Claude Code / Cursor: add the server to your MCP config with the command above
- Registries: listed in the official MCP Registry as
io.github.creatorscore/creatorscore-mcp, on Smithery and Glama
- Hosted connector: Claude and ChatGPT users can add CreatorScore from creatorscore.io/mcp without installing anything
What it can do
help | Show the menu of everything CreatorScore can do and what each option needs. | Free |
check_account | Show your credit balance and per-platform pricing. | Free |
vet_creator | Vet one creator across all their accounts at once (you supply the handles) — unified cross-platform CreatorScore + per-platform breakdown + worst risks anywhere. Never scrapes. | 1 credit per scored account |
check_score_job | Poll a fresh-scoring job until it's done. | Free |
get_creator_report | The evidence layer for a scored creator: every active risk flag with its proof (post link, caption/transcript excerpt, on-screen text, reviewer reasoning, archived thumbnail + flagged-frame links), brand mentions with a disclosure verdict per brand (undisclosed partnerships = FTC exposure), controversy + web findings, SHAP drivers per agent, audience, following-graph risk, score history — plus a browser-openable PDF link. Whole footprint from one handle. Never scrapes. | Included for a creator you already paid to score; 1 credit as a cached read otherwise |
start_background_check / check_background_check / list_background_checks | Order, poll, and list Social Background Checks — a separate brand-safety inspection (no score). Completed checks return the categorized findings with evidence and a PDF link. | $19.99 one-time via Stripe to order; reads are free |
calculate_fair_rate | OCRS rate calculator — what to charge/pay for a sponsored post. Returns a fair rate + confidence band + multiplier breakdown. Pure math, no scraping or scoring. | Free (0 credits) |
Safety by design
- Footprint, not single accounts. A CreatorScore is always a creator's whole cross-platform footprint, never one isolated account. The scoring tools read cached scores first and never silently scrape — triggering a fresh (paid) scoring pass is a separate, explicit,
confirm-gated tool.
- Transparency. Score output always names the what behind a penalty — every knockout shows its reason, every risk flag shows its description. Never "2 issues detected" with no detail.
Setup
1. Get an API key
Sign up at creatorscore.io/signup/developer or grab a key from creatorscore.io/account/api. Buy a credit pack from the same page.
2. Add it to your MCP client
No install step — npx fetches and runs the published package on demand.
Claude Desktop — edit claude_desktop_config.json:
{
"mcpServers": {
"creatorscore": {
"command": "npx",
"args": ["-y", "creatorscore-mcp"],
"env": {
"CREATORSCORE_API_KEY": "sk_live_..."
}
}
}
}
Claude Code — one command:
claude mcp add creatorscore -e CREATORSCORE_API_KEY=sk_live_... -- npx -y creatorscore-mcp
Then restart the client. Try: "Use creatorscore to check my account balance, then vet @mrbeast on youtube."
Working from a clone of this repo instead? You can run it from source: npm install && npm run build in mcp-server/, then point command/args at node /absolute/path/to/mcp-server/dist/index.js.
Configuration
CREATORSCORE_API_KEY | ✅ | — | Your X-API-Key. |
CREATORSCORE_API_BASE | | https://creatorscore.io/api/v1 | Override the API base URL (e.g. for staging). |
How a fresh score flows
score_creator_fresh (confirm: true)
│ POST-style GET /api/v1/creators/{platform}/{handle}
▼
CreatorScore servers: scrape → transcribe every video → v2.2 scoring
│ returns 202 + job_id
▼
check_score_job (poll ~every 30s) ──► status: done → full score
The agent never touches the scrapers, the ML service, or any provider key. It asks the API for a score; the API does the work behind the credit wall.
License
UNLICENSED — internal CreatorScore project.