Give the GPT Action a Memory with History

By default each AI step answers with no knowledge of what came before. Turning on History lets the GPT action remember earlier exchanges, so replies stay consistent and personal across steps or runs.

History types

Choose how widely the memory is shared:

• This Account: shared by every workflow that uses this history type in your account.
• This Workflow: shared by all AI steps in one workflow.
• Per Execution: remembers only within a single run for one contact.
• This Step: remembers across many runs of one specific action.
• Custom: a named memory you create and reuse across steps or workflows.

History is kept separately for each contact.

Turn it on

1. Enable the feature in Labs if it is not already on.
2. Add or open a GPT powered by OpenAI action and set the type to Custom.
3. Switch on Enable History.
4. Choose a value for History for.
5. Add System Instructions, if wanted.
6. Save.

History applies to the Custom action type only. When it is on, a default model is selected for you.

Advanced options

• System Instructions: rules the AI follows every time, such as tone or required format.
• Exclude instructions from history: keeps these rules out of the stored memory.
• Exclude responses from history: stops saving the AI's answers, which suits one-word sentiment results or analytics outputs.

Example

An onboarding workflow asks a new client three questions across several days. With Per Execution history, the final summary step knows all the earlier answers without you pasting them in.

Good to know: Choose the narrowest history type that does the job; a wide memory can carry old, irrelevant context into new replies.