How to control Junie's quota usage to avoid high token consumption?

Situation

Our team recommends using AI Ultimate with Junie to benefit from significantly bigger quota numbers. This said, an additional solution is to try and control Junie's usage better.

Question

How to better control Junie's usage / quota limits?

Answer

Here are some recommendations that may help to better control Junie’s usage:

• Use the Default AI model for Junie. According to our benchmarks, this model provides the best quality and it's considerably cheaper than alternatives. Currently the default model is Gemini 3 Flash.

  

• Break large tasks into smaller ones

  Smaller and concrete tasks normally take fewer tokens because Junie requires less context exploration, less requests to LLMs.

• Use a compact, structured prompt per task

  Instead of long, free-form prompts, use a short structure like: Goal: Implement X. Scope: Only file A and file B. Don’t touch other modules. Edits: Small changes only, no refactors, no new files, no renames. Questions: Ask before doing anything outside this scope.

  This keeps prompts small while preventing unsolicited refactors, “vibe coding”, or Junie trying to drive the project.

• Use follow-ups and "Think more" wisely

  Follow-ups and improvements, as well as the "Think more" option with Junie may take more tokens. To minimize this, be sure to create clear instructions for Junie from the beginning and add them to the guidelines.

• Use Junie and AI Assistant in the same project but for different tasks

  Use Junie (agent) for multi-step implementation work. For simple questions that don’t require planning or autonomous edits, use AI Assistant Chat / Ask mode instead – it’s cheaper and doesn’t change your code unless you explicitly apply suggestions.

• Keep your guidelines file concise

  If you use a guidelines file (.junie/AGENTS.md), keep it short and clean to reduce quota usage. To reduce quota and prevent unwanted behavior:

  • Keep the file short (around 20–40 lines).

  • Focus on rules and constraints instead of long explanations.

  • Add explicit “forbidden behavior” rules, for example:

    - Do NOT refactor or clean up code unless I explicitly say “refactor”.

    - Do NOT rename files, move code, or change project structure.

    - Do NOT modify .junie/guidelines.md or other configuration files.

    - Ask before changing more than 3 files in one task.

See also

How to control AI Assistant's quota usage to avoid high token consumption?