But you know there are free tools that do what this MCP server does already, right? And a user could also just point the agent at the Data API for more flexibility…
Yes, there are free tools in this space and we are aware of them. A few things we offer that the free tools we have seen currently do not:
- Local installation via .mcpb — your credentials never leave your machine and never touch a third party server
- Multiple Bubble apps in one Claude session without switching tools
- Write operations: create, update, delete records and trigger workflows, as well as read records, from the same tool
- Packaged prompts for schema analysis, test data generation, and orphaned record detection
On pointing an agent directly at the Data API - absolutely possible for someone comfortable building and maintaining their own integration. Our tool (and yours I believe) is for teams who would rather not.
As for free: we are a two person team based in Spain building and maintaining this tool. Paid is what makes that sustainable and makes sure we are there when something breaks, there’s an API update or users come up with a new feature that will bring them more value.
In my own project, I have a parallel database built for AI analytics. It’s built with the principle of fast, cheap inference.
Your DB setup should be built around what services they serve. That can be well built schemas in Bubble, or in external infra. You can alternatively use services like Databricks or Cloudflare’s AI Search where they already have pipelines to automate making your data better for AI inference.
The point of services like these are the same points I am trying to make: You can’t point AI to your raw data and not expect degradation and exponential costs.
hey @ihsanzainal84 totally agree that for reporting purposes, pulling all data live via the API at once is a dead end. It needs to be cached either locally or in a remote database.
Coming back to the original discussion - there are 2 separate goals here:
- Extracting Bubble data for analysis and reporting (needs local or remote database)
- Handling operations in your Bubble app to speed up workflows (create new records, search for individual records - all in the context of connecting to other apps and orchestrated by AI)
I would be interested to hear of any use cases for goal #2
Wishing you luck! But, personally, doesn’t make sense. The costs as everyone stated aren’t worth the extras provided to any Bubbler I know.
MCP is visible but known for its inconsistent performance, compared to CLI, a lot of services stopped marketing their MCP for that reason and instead talk about installing and integrating their CLI with Claude or Codex.