MCP for AI agents
Model Context Protocol endpoint for Cursor, Claude Desktop, and custom agents. MCP is a thin adapter over the same core services as the REST API — business logic lives in one place.
Related: Developer hub · REST API · Webhooks
What you get on paid plans
- AI contest assistant — chat-based builder in the dashboard
- MCP server — list contests, stats, CSV exports via standard MCP clients
- REST API + webhooks — same account keys for Make.com and Zapier
Free accounts can run unlimited contests; upgrade for assistant, MCP, and integrations. Compare plans.
Setup
- Sign in to the dashboard and subscribe to any paid plan.
- Open Account → API keys & integrations and generate a key.
- Add the MCP server to your agent (Cursor example below).
Endpoints
| URL | Auth | Purpose |
|---|---|---|
https://app.rewardsfuel.com/api/mcp | Bearer rf_live_… | Full MCP server (paid tools) |
https://app.rewardsfuel.com/api/mcp/public | None | Free draw_winners only |
API key needs mcp scope (included on new keys). Write tools also need write scope.
Cursor example
{
"mcpServers": {
"rewards-fuel": {
"url": "https://app.rewardsfuel.com/api/mcp",
"headers": {
"Authorization": "Bearer rf_live_YOUR_KEY"
}
}
}
}
Paid tools — read
list_contests — accessible contests (owned + shared) with entry counts get_contest — details and entry methods get_contest_stats — totals and top contestants export_contestants_csv — unique contestants as CSV text export_entries_csv — all entry rows as CSV text
Paid tools — write (write scope)
create_contest — blank or full draft create update_contest — patch fields or status sync_contest — full draft reconcile create_entry_method / update_entry_method / delete_entry_method create_prize / update_prize / delete_prize
Free tool
draw_winners — verifiable random draw saved to rfwin.uk (all accounts; also on /api/mcp/public)
Protocol
JSON-RPC 2.0 over HTTP (POST). Supported methods:
initialize tools/list tools/call ping
GET /api/mcp returns server metadata (no auth required for discovery).
Security
- Token-scoped authz: account scope comes from your API key, never from tool arguments. Agents cannot pass
accountId to access another tenant. - Least privilege: use
read + hooks for export-only agents; add write only when agents must configure contests. - Rate limit: 60 tool calls/minute per API key.
- Audit: MCP calls are logged server-side by account and key.
REST vs MCP vs webhooks
Integration Best for Webhooks Make/Zapier real-time triggers (new entrant, winner picked) REST v1 Scheduled exports, polling, custom backends, HTTP mutations with write MCP Cursor, Claude, and other AI agents
Questions? Contact us.