Hi All!
First post here, so let me know if I’m in the wrong spot. I’ve been working through my first Coda pack for our VoIP service provider Dialpad.com although its pretty straightforward, its far from “simple” or “low-code” like everything else Coda so far.
I’m no typescript/javascript expert (most of my coding experience was a decade ago in college), but I’m just capable enough to do some copy pasting and error troubleshooting going back and forth between Coda’s SDK and the Dialpad API documentation. Since I’m able to figure it out by literally mimicking examples and other packs, I’ve got to believe this should be able to be automated with some type of AI tool that simply ingests the target app’s API documentation and spits out the Coda pack code.
@Scott_Collier-Weir, I saw your previous post about writing schema using ChatGPT - curious if you’ve developed that further in recent years, or write it all from scratch now?
With the direction Coda/Grammarly seem to be headed leveraging these API connector packs (or whatever other framework they are building for Coda brain, Grammarly etc), it would be awesome to see some type of native tool within the Coda pack editor that can basically autogenerate packs by ingesting API documentation. @Eric_Koleda - any chance that’s on the roadmap for packs in the future?
Hi @Tyler_Maskiewicz - Happy to hear that you’re dabbling in Pack making, and those college coding skills are being dusted off! I’ve talked to a number of folks that have built Packs using AI, and zooming out there is definitely a rising trend of “vibe coding” and new AI coding tools.
We currently don’t have any plans to integrate AI into the Pack Studio directly, but it’s a feature request we have gotten in the past. In the mean time I would encourage you to try some of the AI tools out there and let us know how it goes!
2 Likes
Is anyone else done completely done with vibe coding yet? I’m building at the moment with Cursor and RepoPrompt - I’ve spent three weeks putting together my skellington with unit testing, e2e, integrations tests… getting it deployed and having a solid foundation to actually build some functionality… only for somebody to post on LinkedIn something they vibe coded on Next.JS and managed to deployed to Vercel… supposedly in just two evenings?
One thing I would say about AI is that Coda needs to think about MCP and how that’s going to play out? Pipedream are already offering MCP to Coda.
I’d add that I don’t have a use-case for MCP yet - but out of the AI hype train, MCP doesnt look like its going away
1 Like
@Gareth_Simpson1 - thanks for sharing about pipedream’s MCP for CODA. I think that’s sort of what I am getting at. At some point, me writing the Coda pack is just me manually translating between two sets of developer documentation - a slow manual process for the unlearned, but a manual process even for the learned. Eventually, I could see something like an llm and MCP making it easier to develop these packs providing more “infinite” opportunity assuming all the “tools” provide MCP servers. Presently, I suppose most companies large enough to provide those MCP servers likely already has an existing Coda Pack developed.