Feature
An AI assistant that lives in the editor.
Writing webhook handlers is repetitive. So is adding error handling, parsing request bodies, and remembering the exact shape of a Stripe event. The nvoke AI assistant does the boring parts so you can focus on the business logic.
What the assistant actually does
The assistant lives in a panel next to your code. It has your current function in context, so its suggestions are about your code, not a generic example from a training dataset. Ask it to add input validation and it writes the validation for the shape of the request body your function is already handling.
When it returns code, there is an Apply button. Click it, and the editor updates with the new code. You can undo, edit, or keep iterating with a follow-up prompt. The assistant is not a black box that ships to production — you still read the code, hit save, and watch the log stream.
The prompts that do real work
A few starter prompts live at the bottom of the panel because they are what people ask for most often:
"Write a hello world function." For when you are spinning up a new function and want a working starting point in two seconds.
"Add error handling to this function." Wraps the existing body in a try/catch with sensible status codes and a useful response shape. The assistant reads what can throw and handles it specifically, not with a generic return 500.
"Explain this code." Useful when you are reading a function you wrote three months ago. Also useful when you are reading a function someone else wrote and want a quick summary before you change it.
"Turn this into a Stripe webhook." Takes a skeleton function and adds signature verification, event type routing, and error handling wired for Stripe's retry behavior. You still need to add your own business logic inside the event handlers.
Where it fits in the workflow
The assistant is fastest at the start of a function and at the end of a refactor. At the start, it gets you from a blank editor to a running endpoint in a conversation. At the end, it is good at the last-mile polish: adding types, tightening error messages, writing the boring parts of a test.
It is not good at — and should not be used for — anything you cannot verify with a quick read. Every line it writes ends up on your endpoint. Treat the assistant like a junior pair: useful, fast, occasionally wrong.
Context-aware
The assistant sees the function you are editing. Ask "add error handling" and it rewrites your actual code, not a generic example.
One-click apply
When the assistant returns code, a single click drops it into the editor. No copy-paste shuffle, no reformatting.
Starter prompts
Common tasks are one click away: hello world, error handling, code explanation, or "turn this into a Stripe webhook".
Try the assistant.
Open a function in the editor and hit the sparkles icon. The first prompt is free, and so are the next hundred.