Looking for a best-practice that works for n8n teams working in collaboration with clients:
Working together with clients/designers/product-owners on a n8n automation, especially when it comes to AI Agents with prompts, means a lot of back and forth over the prompts.
Modifying the workflow could mean you need to adapt the prompts, and vice-versa modifying the prompts often requires adapting the workflow (e.g. to accommodate a new expected structured output).
n8n doesn’t allow simultaneous editing of a workflow, so work has to be done linearly.
Since a workflow is a living creature (bugs are found, new requirements arise etc), there is a need for a CI-like approach (Continuous Integration) where automation developer and product owner can work independently on prompts without breaking the automation and without fist-fighting over the prompt ![]()
The solution I came up with for now is far from ideal:
- We manage the prompts in promptlayer.com , and excellent way to collaborate over prompts, with versioning and many other tools.
- I then use a sub-workflow like the one below, to pull prompts and assign them to the corresponding agent.
This allows the client to develop the prompts in parallel to my work on the automation, but it’s far from being ideal as it means a lot of communication around what was changed on one side and what needs to be adapted on the other side.
Has anyone found a viable solution for this problem?
Is there a better way to approach this?