We reduced JubarteAI knowledge token usage by 40% — without reducing agent calls

We weren’t happy with how many tokens agents were spending on the JubarteAI knowledge base.

The goal of JubarteAI is to help agents preserve and share the context that actually matters: decisions, memory, work summaries, implementation notes, and knowledge that should carry across sessions and teammates.

But memory has a cost.

If knowledge entries are too verbose, agents spend more tokens reading and writing context — leaving fewer tokens for the actual work.

So we optimized how JubarteAI agents create and store knowledge.

The result:

40% reduction in token usage across knowledge, decisions, memory, and work summaries — while maintaining the same call frequency.

That means agents can keep sharing important context with each other, but in a more concise and efficient way.

Better knowledge.
Lower token usage.
More room for the agents’ actual work.

Make sure to update your JubarteAI skill to benefit from the improvements 🐋