Limitations
This page focuses on MCP tools (names like generate_media, get_models). REST API routes use similar ideas but different path names and some different rate buckets (e.g. POST /v1/upload/media); see REST API.
#Rate limits
Requests are limited per minute per account (or per connected client token). Approximate limits:
| Tool or category | Limit (requests per minute) |
|---|---|
| generate_media | 30 |
| generate_music | 10 |
| generate_dialogue | 10 (separate bucket) |
| create_ad_copy | 5 |
| get_generation_status, get_music_status, get_models, get_balance, get_generations, get_generation_estimate | 120 |
If you exceed the limit, the server returns a rate limited error. Wait a short time (e.g. until the next minute) before retrying. Your client may show a message like “Too many requests” or “Rate limit exceeded.”
#Credits
- All generations (media, music, ads) consume Kubeez credits from your account.
- Cost depends on the model and options (e.g. quality, duration). Use get_models to see cost per generation for each model.
- Before starting a job, use get_balance to ensure you have enough credits. If you don’t, the request may fail with an insufficient-credits error.
- You can purchase or top up credits in your Kubeez account.
#Errors
- Rate limited - Too many requests in a short time. Retry after a brief wait.
- Insufficient credits - Your balance is too low for the requested operation. Add credits in your Kubeez account.
- Invalid or expired token - Re-authorize the MCP client from Kubeez (e.g. disconnect and connect again).
- Invalid parameters - e.g. wrong model for the chosen generation type, or missing input image when the model requires one. Check get_models and the tool docs (e.g. Media tools).
#Tool-specific limits
#generate_dialogue
- Minimum characters: 5 characters total.
- Maximum characters: 5,000 characters total per generation.
- Concurrent jobs: Up to 5 generations at once.
#generate_media
- Concurrent jobs: Up to 10 generations at once.
#generate_music
- Concurrent jobs: Up to 5 generations at once.
When a generation fails, the status response usually includes an error message describing what went wrong.
See How it works for recommended workflows and retry behavior.