Is “Auto” mode unlimited and free?

Support Help
1

Hi everyone,

I have a quick question about Auto mode.

Is Auto mode unlimited, or are there any usage limits? Also, does it incur any additional monthly charges, or is it fully included in the subscription?

I couldn’t find a clear explanation about this in the documentation, so I’d appreciate it if someone could clarify.

Thanks in advance!

3

Auto mode isn’t unlimited, it just counts toward whatever usage limits your plan has. Auto mode is choosing the best model for the task. If you hit your plan’s limit, you’ll typically be rate-limited or prompted to upgrade rather than automatically billed more. If you’re on a team or enterprise plan, limits may be shared across the workspace.

5

I don’t quite understand. May I ask if exceeding the traffic limit means slower replies. But is the number of questions actually infinite?

6

hi @Lucas_Cavalheri_TH & @Sounean , thanks for the great questions!

We do have two usage pools but we do not use rate-limits.

  • Auto: This includes Auto mode as well as Composer 1.5 and does not count towards your limit for other models. This way you get extra usage of Auto / Composer 1.5.
  • API: This is the pool for all other models like GPT-5.4, Opus, Sonnet, Gemini,… and is based on your plan e.g. $20 worth of AI model usage.

Both the Auto / Composer & API pools are visible in your Usage dashboard or in Cursor Settings.

Once you reach the limit on the API usage pool you will see several options:

  • Upgrade plan: starts a new plan with higher usage limit e.g. $60.
  • Enable On Demand: starts separately billed usage and you can set the limit. e.g. $50
  • Switch To Auto: Continues to use Auto mode

The actual usage is counted based on AI models token consumption and that models cost as reported by AI providers when you make requests to Agent.

  • Larger models cost more per token than smaller ones, but they are more capable
  • Thinking models can reason through the task, which also uses more tokens. than non thinking models.

Overall the plans are not unlimited, but we do not slow down replies or reduce quality of output.