Glossary
Working memory
Working memory is what an agent holds while it works on the current task. It is the small set of context the model can see at one moment.
What working memory is
Working memory is the context an AI agent keeps in view while it handles a task. It includes the instructions, the current request and the facts it needs right now.
It is short-lived. It is rebuilt for each step or turn, and it is not the same as what the agent stores long term.
Why it matters
- Space is limited. Only so much text fits in one prompt.
- Too much irrelevant context makes answers worse. The right facts must be picked, not just added.
How it works in Harny
- Harny gives each person a private memory the app's AI can read.
- Your server calls
POST /v1/memory/recallwith a query. SetlevelormaxTokensto cap how many tokens the results use.
Is working memory the same as memory?
No. Working memory is what the agent sees during one task. Memory is what it can recall across tasks.
Keep reading
Go on. Get Harny.
The first 500 monthly-active end users are free for good. Create an app key and make your first call in minutes.