From Visual Workflows to Native Code in Production: The Complete Journey of an n8n Backend That Couldn't Stop Evolving

In the early days, the author built an auto repair shop management SaaS entirely using n8n workflows, from AI customer service to yard management. However, as the product grew and required staging and production environments, the synchronization problem between these environments emerged. Since workflows existed only in the server's database without a file history, there was no way to track changes, who made them, or when.

Exporting and importing workflows manually between servers was impractical, especially with 74 workflows in place. The author needed a solution to capture the current state of workflows as files, version them in Git, and automate deployment to the production server. He built GitOps for n8n, turning visual workflows into versionable code using n8n's REST API.

This GitOps structure allowed for easy synchronization, review, and deployment of workflows between staging and production. However, as the product continued to grow, the cost of running n8n instances for 126 workflows became evident. Each n8n node serialized and deserialized the entire state, resulting in high latency and memory usage.

The author considered two options: rewriting the backend in native code or scaling horizontally with more n8n instances. Both options were deemed undesirable. The turning point came when the author realized he already had 74 workflows as structured JSON files in the Git repository. With this realization, he could now treat the workflows as structured code, enabling further automation and improving the overall speed and scalability of the backend.