Agents Memory

Agents Memory is what agents recall about your Projects. It keeps a record of the Experiments Discovery Runs have concluded, across all your Projects, so past results are not lost between runs and agents can draw on what has already been tried.

Stored history is a record of past Experiments. It is not by itself evidence that later runs improve.

Accessing Agents Memory

Open the More menu (the … button in the sidebar) and choose Agents Memory.

The page lists your Memories, newest first. Each row shows the Experiment's title, the kind of Memory and the Project it came from. Memories appear here as Experiments are validated in Discovery Runs.

A single Memory

Click a Memory to open its details:

• Hypothesis: what the Experiment set out to test.
• Conclusion: how it was concluded, for example Validated.
• Confidence: the reviewing models' confidence in the Experiment. This is a model judgement, not a statistical measure.
• Created: when the Experiment was created.
• Versions: the Versions it produced.

Open in Discovery takes you to the Experiment in its Discovery Run.

Data boundaries

Memories stay inside your Artemis deployment.