# Tunova > Prepaid REST API + MCP server for Suno AI music generation (models v6, v5.5). > You pay only for songs that actually render: 1 song = 10 tokens, lyrics = 2 tokens; > a failed render auto-refunds. Auth is a single `X-API-Key` header. 50 free tokens on > signup, no card. This host (api.tunova.ai) is the API; sign up + billing live at https://tunova.ai. ## Docs - [API reference](https://api.tunova.ai/docs): endpoints, auth, errors, HMAC-signed webhooks - [OpenAPI 3.1 spec](https://api.tunova.ai/openapi.json): machine-readable, the source of truth - [MCP server](https://api.tunova.ai/mcp): POST JSON-RPC (streamable-http); tools generate_song, wait_for_song, check_song; same `X-API-Key` - [Authentication](https://api.tunova.ai/.well-known/auth.md): how to get and use a key ## Workflow - POST /api/generate `{"prompt":"..."}` (optional `model`: v6 | v5.5) → 202 `{job_id, status_url}`. - Use `model: "v6"` for Suno v6. `v5.5` remains a compatibility alias for the same v6 model, at the same cost; omitted-model requests keep their existing default. - Read the current partner token balance with GET /api/balance using the same X-API-Key. - Poll GET /api/jobs/{job_id} every ~10s (a render takes ~1-3 min) until `status: "complete"` → `clips[].audio_url`. `status: "failed"` refunds the 10 tokens automatically. - Custom lyrics: POST /api/custom_generate `{"prompt": , "tags":