Memory

Memory is what the agent knows about reaching things in your project: where a page moved, which email is slow, what to wait for, what to avoid. The agent uses it as hints when it works out how to reach a step's outcome.

Memory can be wrong, and it's always shown. It never adds steps, never changes a result, and never records a bug. Bugs are issues; expected results are steps of a test case.

Open a project and click Memory.

Product knowledge

Product knowledge is a briefing you write for the agent, like one for a new tester: what the product does, who uses it, words it uses, and anything a careful tester should know. It's given to the agent in every run, on every environment, and it's also used when Autonomy drafts steps from a description. It isn't learned from runs.

Memories

A memory has a title and content, such as:

OTP emails are slow on Staging Wait a minute and re-read the inbox before asking for a new code.

Each memory has a category (Site structure, Test insights or Preferences & rules), an importance, and where it applies: All environments, or one environment.

Click Add memory to write one. Autonomy checks a memory before saving it, and refuses one that records a bug or an expected result: "This is a memory about how the app works. Record bugs as Issues instead, and expected results as steps of a test case."

Learned from runs

After each run, the agent can suggest what it learned, such as a page that moved or an email that took a while. Suggestions appear under Suggested from runs, with the run they came from. Click Keep, Edit or Dismiss. Kept suggestions are used from the next run.

Settings

Owners and admins can change these switches, which are all on by default:

An archived project learns nothing.