Auditable belief history
What did the agent believe when it acted — and what has changed since? Answerable, because nothing is edited in place.
observesupersedeWithout Klio
When an agent’s action is questioned, most memory systems can only show you what they believe now. The belief that actually drove the action is gone — overwritten by its successor.
With Klio in the loop
Every memory is a dated, attributed row; changes of mind are new rows pointing back at what they replaced.
Provenance marks what was stated by a person versus inferred by a model.
To reconstruct a moment, walk the supersession chain back to it.
The call an agent makes
explore(
{
"entity": "deploy policy"
}
)Real tool, illustrative payload. The full contract for every tool is in the docs.
What changes
Reviews get evidence instead of recollection. The record shows the belief, its source, and its retirement — with dates.
Related use cases
Free and unlimited for one person. Connect an agent and try this walkthrough on your own project.