Hi all,
I've been exploring how far AI tools like Claude can go in actually developing Celigo integrations, not just assisting here and there. The goal: fewer manual touches, more end-to-end automation.
Curious how others are approaching this:
-
Have you been able to get AI to scaffold full flows (exports, imports, mappings, hooks) with minimal back-and-forth?
-
How are you feeding it context, like connector docs, schemas, or sample payloads, so it produces usable output the first time?
-
Any luck chaining AI with Celigo's APIs to auto-generate or update integration components?
-
Where does it still break down and force you to step in manually?
Trying to figure out a workflow where AI handles most of the heavy lifting, and I mainly review and adjust. Would love to hear what's working (or not) for others.
Cheers!
Hi @Senal_Punsara, congrats on your first post on Connective!
We have been doing plenty of experimentation with the API and have found that LLMs can be somewhat successful by brute force. But there is a better way...
We have not broadly announced it yet, but please check out https://developer.celigo.com, which describes our new CLI. The CLI wraps the API (obviously) but provides additional context to the LLM that makes life a lot easier. We've seen improvements on all fronts -- cheaper, faster, better.
We made it available to partners as part of a closed Beta in April, and we will announce it more formally shortly, but you can use it today. I recommend you proceed with some caution -- eg: use the CLI to write to a non-prod environment, and then switch to the UI (and Ora) for validation.
Let us know how you go!
Cheers,
Matt