To simplify the development process, we created a very basic CLI which creates boilerplate code to get started, builds the node (as they are written in TypeScript), and copies it to the correct location.
That command will build the JavaScript version of the node from the TypeScript code and copy it to the user folder where custom nodes get read from `~/.n8n/custom/`
1. Restart n8n and refresh the window so that the new node gets displayed
Never change the incoming data a node receives (which can be queried with `this.getInputData()`) as it gets shared by all nodes. If data has to get added, changed or deleted it has to be cloned and the new data returned. If that is not done, sibling nodes which execute after the current one will operate on the altered data and would process different data than they were supposed to.
It is however not needed to always clone all the data. If a node for, example only, changes only the binary data but not the JSON data, a new item can be created which reuses the reference to the JSON item.
An example can be seen in the code of the [ReadBinartFile-Node](https://github.com/n8n-io/n8n/blob/master/packages/nodes-base/nodes/ReadBinaryFile.node.ts#L69-L83).
All code of n8n is written in TypeScript and hence, the nodes should also be written in TypeScript. That makes development easier, faster, and avoids at least some bugs.
Some third-party services have their own libraries on npm which make it easier to create an integration. It can be quite tempting to use them. The problem with those is that you add another dependency and not just one you add but also all the dependencies of the dependencies. This means more and more code gets added, has to get loaded, can introduce security vulnerabilities, bugs and so on. So please use the built-in module which can be used like this:
That is simply using the npm package [`request-promise-native`](https://github.com/request/request-promise-native) which is the basic npm `request` module but with promises. For a full set of `options` consider looking at [the underlying `request` options documentation](https://github.com/request/request#requestoptions-callback).
When a node can perform multiple operations like edit and delete some kind of entity, for both operations, it would need an entity-id. Do not call them "editId" and "deleteId" simply call them "id". n8n can handle multiple parameters with the same name without a problem as long as only one is visible. To make sure that is the case, the "displayOptions" can be used. By keeping the same name, the value can be kept if a user switches the operation from "edit" to "delete".
Some nodes may need a lot of options. Add only the very important ones to the top level and for all others, create an "Options" parameter where they can be added if needed. This ensures that the interface stays clean and does not unnecessarily confuse people. A good example of that would be the XML node.
There is not much of a guideline yet but if your node can do multiple things, call the parameter which sets the behavior either "mode" (like "Merge" and "XML" node) or "operation" like the most other ones. If these operations can be done on different resources (like "User" or "Order) create a "resource" parameter (like "Pipedrive" and "Trello" node)