Just saw the announcement about Relevance AI's new free tier. 100 runs/month is a solid move to get people in the door. It's enough to wire up a test workflow and kick the tires on their AI agents and vector search.
But my immediate, pragmatic question is: what's the realistic path from there? If I'm prototyping a customer support agent that checks an order database or a tool to auto-categorize support tickets, 100 runs gets burned through in a couple of days of serious testing. Then you hit the paywall.
My main concerns are about the jump to paid:
* **Cost predictability:** Their Pro plan starts at $199/month. That's a steep leap from $0. For a simple, low-volume workflow, that's middleware overkill.
* **Workflow continuity:** If I build a prototype on the free tier that my team starts to rely on for even minor tasks, hitting the run limit means everything just stops. That's a brittle connection, which I hate.
* **Architecture lock-in:** Their platform looks built for chaining multiple AI steps. If I design around that and then find the scaling cost prohibitive, extracting the logic to a more basic API setup (like direct OpenAI + a queue) becomes a rewrite.
For comparison, a simple middleware workflow in Celigo or Workato, while not AI-native, gives me predictable, per-connection pricing and can call external AI APIs with my own keys. The trade-off is more configuration.
Has anyone here actually stress-tested moving a prototype from their free tier into a production workflow? I'm less interested in the AI features themselves and more in how they handle the integration and scaling piece—webhooks, error handling, monitoring. That's where these platforms usually show their true colors.
Integration is not a project, it's a lifestyle.