What are you doing to reduce token usage in Devin ?
Our company has a fixed monthly ACU budget, so using high-end models regularly means I run out of credits within a week. We also can't use most open-source models (DeepSeek, GLM, Kimi, etc.) because of security restrictions they are blocked.
I'm looking for practical ways to reduce token usage while using Devin. Are you using any extensions, GitHub repos, or workflows that help minimize token consumption without reducing output quality?
Any tips or best practices would be appreciated.