The idea is:
Add an option to Execute Step that reuses the latest successful manual execution data instead of re-running all upstream nodes.
When a developer adds a new node after a previously executed node, clicking Execute Step should use the existing output from the previous execution if the upstream nodes have not changed. This would avoid unnecessary API calls during development.
My use case:
I build large AI workflows in n8n Cloud that use OpenAI and other paid APIs.
A typical workflow looks like this:
Trigger → HTTP Request → OpenAI → Code → New Node
After the workflow has already executed successfully, I often add a new node after the Code node. When I click Execute Step to test only that new node, n8n re-executes the HTTP Request and OpenAI nodes instead of using the existing execution data.
This results in unnecessary API calls, additional costs, and much slower development.
I think it would be beneficial to add this because:
It would significantly reduce development costs for workflows using paid AI services.
It would improve the developer experience when building large workflows.
It would make debugging and iterative development much faster.
It would reduce unnecessary API traffic.
It would be especially valuable for OpenAI, Anthropic, Gemini, and other paid integrations.
This could be implemented as an optional setting, for example:
“Reuse latest execution data”
“Execute using previous manual execution”
"Do not re-run unchanged upstream nodes"Any resources to support this?
Several users have discussed similar behavior in the n8n Community Forum, particularly when working with OpenAI and other paid API nodes during manual testing. This feature would address a common pain point for developers building AI workflows.
Unfortunately, no. I am submitting this as a feature request because it would greatly improve the development experience for AI workflows in n8n Cloud.