Memory is what the agent learns over time. As you work with Kiro Web and give it feedback, it picks up your preferences and applies them to future work, so you don't have to restate the same guidance in every session.
You can see what Kiro has learned in the Memory section of your Kiro Web Settings. Memory is maintained automatically, so there's nothing to turn on and nothing to add manually. You can delete any memory you don't want the agent to keep. Until Kiro has collected anything, this section shows No memories yet.
The most direct way to influence what Kiro learns is to give feedback as you work. When you leave guidance on a pull request, such as "always use our standard error handling" or "follow our naming conventions," the agent learns those patterns and applies them to future work across all your repositories.
Only your feedback, as the user who created the task, influences what the agent learns. Other reviewers' comments don't affect it. The clearer and more consistent your direction, the more useful memory becomes over time.
Both memory and steering help the agent follow your preferences, but you create them in different ways.
| Memory | Steering | |
|---|---|---|
| How it's created | The agent learns it automatically as you work and give feedback | You write it explicitly, in Settings or as markdown files in .kiro/steering/ |
| Where it lives | In Settings | In Settings and in the filesystem |
| User can control | Deletions | All edits |
Use steering when you want to state a rule explicitly and control exactly what the agent follows. Memory complements it by capturing the preferences the agent picks up as you work together.
Memory