Token Optimization
Token Optimization helps the AI agent complete tasks more efficiently. It reduces unnecessary tool calls and actions, improves the order of execution, and limits the amount of terminal output sent back to the agent. This lowers token usage and preserves more of the model's context for the work that matters.
How it works
Token Optimization works at two complementary levels. Basic reduces the token cost of tool results. Max also helps the agent choose a more efficient path through the task. Both modes preserve the original intent of your request and fall back to standard behavior when an optimization is not appropriate.
Basic and Max modes
Basic
Basic focuses on the tokens used by tool calls. It reduces noisy or repetitive command output before it is added to the conversation, while keeping the information the agent needs to continue working.
The agent otherwise follows its normal approach to the task. This makes Basic a predictable default for everyday work, especially sessions with frequent terminal commands or verbose output.
Max
Max includes everything in Basic and additionally optimizes the agent's working strategy. It encourages a better order of operations, groups related work, and reduces redundant investigation, repeated calls, and avoidable rework.
Max is best suited to longer or multi-step tasks where planning and tool-use decisions can have a larger impact on total token usage. It does not change your goal or remove required validation.
Choosing a mode
- Choose Basic for a conservative reduction in tool-output tokens without changing how the agent approaches the task.
- Choose Max when you want the largest overall reduction across tool output, tool calls, and unnecessary work.
- Choose Off when you need the host's unmodified behavior and complete original command output.
Actual savings depend on the task, tools, and amount of generated output. Optimization is designed to reduce waste, not to skip requested work or required checks.
Enable Token Optimization
- Open the Supercode menu and use the Token Optimization switch.
- Alternatively, select Settings in the same menu. On the General page, find Command Optimization and use the Token optimization switch.
- The Settings page shows installation, update, and retry status when action is required.