Using Agent Learning to Help Your Agent Remember Past Runs

What is Agent Learning?

Agent Learning is an optional memory feature for GoMarble Agents. When turned on for a specific agent, that agent's scheduled runs record what they found and did. Future runs can then reference that recent history to avoid repeating themselves — for example, saying "still open, day 3" instead of flagging the same issue as new, or skipping a connection it already knows is broken.

Agent Learning is off by default. Nothing is stored or read from past runs unless you explicitly turn it on for a given agent.

Why is it off by default?

Storing run history is a deliberate choice that should belong to you. Because past run data is written and read only when this toggle is enabled, turning it on is your explicit consent for that agent to use memory. If you never turn it on, the agent behaves exactly as it always has — each run starts fresh with no knowledge of previous runs.

What gets stored?

When Agent Learning is enabled, the agent records the key findings and actions from each scheduled run. That history is then made available to the agent at the start of its next run so it can provide continuity in its analysis and recommendations. No run history is written or read for agents where the toggle remains off.

How to enable Agent Learning

When creating a new agent

1. Start creating a new agent as usual.

2. On Step 2 of the creation flow, locate the Agent learning toggle.

3. Switch the toggle on.

4. Finish setting up and save your agent.

For an existing agent

1. Open the agent you want to update.

2. Open the agent's edit drawer.

3. Locate the Agent learning toggle.

4. Switch the toggle on (or off to disable it).

5. Save your changes.

Tip: Hover over the toggle to see a plain-language description of what it does before you decide to enable it.

How to disable Agent Learning

You can turn Agent Learning off at any time by returning to the agent's edit drawer and switching the toggle off. Once disabled, future runs will no longer write to or read from run memory.

Frequently asked questions

Does enabling Agent Learning affect agents I haven't turned it on for?

No. The toggle is per-agent. Enabling it for one agent has no effect on any other agent.

Does turning off Agent Learning delete stored history?

Turning the toggle off stops future runs from writing or reading run memory. If you have questions about data retention, see the articles in the Data Storage and Security section of the help center.

Who can enable or disable Agent Learning?

Any user who has permission to create or edit an agent can control its Agent Learning setting.