-
Notifications
You must be signed in to change notification settings - Fork 99
Description
The Rive agent is not of suitable quality to produce reliable results yet. Signed up for subscription specifically for AI agent access. Submitted a single request to key frame very simple composite vector shape, and nearly all of the initial credits were consumed in single request. Additionally, there were many mistakes that I need to now clean up manually. I suspect the majority of Rive users already have Claude Code or Codex subscriptions and thus it is a better user experience to simply offload inference to the model providers so that users can actually iterate freely. The whole purpose of upgrading to plan with agent features is to save time and increase efficiency. I'd honestly pay more for current paid plan if it had open MCP access, rather than being forced to use the Rive agent--which is much better economic proposition for Rive vs the current alternative.
Given that Rive is likely paying foundation model company API prices for the agent, there's likely not even that much profit being made anyway, and there's a limit to how much people are willing to pay extra for AI features bolted onto non-AI-native products, considering the much greater relative value provided by foundation model company bundled plans.
Rive is a great product at the end of the day, but just know that the AI agent feature is not going to be a huge profit center. As mentioned, as of now, the only reason for me to stay on a paid plan is for export capability. And I would be willing to pay more if I could freely use my Claude Code subscription inside Rive. But there's absolutely no way I'm upping my plan tier or purchasing AI credit packs for unreliable agent runs.