Who It's For

Use it when:

Flows are directly replayable

Flows are directly replayable

Edge cases can be learned and replayed

Edge cases can be learned and replayed

LLM reasoning is only needed after a certain point

LLM reasoning is only needed after a certain point

Butter is well-suited for non-interactive agents fired off to do data entry, computer use, back-office tasks, and more.

As a general rule of thumb, if in principle the task could be done with code (but maybe you don't have time to write out every case), adding Butter to your agent is a good idea.

Don't use it when:

Flows involve creativity or intelligent generation.

Or a user is instructing a new task every time.

Don't use it when

This would be your chat-based creative/coding agents.

We'd love to support these cases, but realistically you'd have a very low cache hit rate.

Contact us to learn what is/isn't worth caching.