# I built a no-code backend that gives AI agents an MCP tool for real business logic

> Source: <https://dev.to/rubens_rafael_700/i-built-a-no-code-backend-that-gives-ai-agents-an-mcp-tool-for-real-business-logic-3ei5>
> Published: 2026-07-08 19:46:18+00:00

I kept running into the same annoyance building agent side projects: giving an agent DB access is easy, but the moment it needs real business logic — check inventory,

process an order, apply a rule — I was back to writing and hosting a small server just for that one action.

So I built **Calamo**: you define objects (Postgres tables) and actions (workflows)

visually, then publish an action as an MCP tool your agent can call directly, or as a

plain REST endpoint if you'd rather call it from regular code.

[https://www.youtube.com/watch?v=6l4rlK88wrg](https://www.youtube.com/watch?v=6l4rlK88wrg)

In the video: I create an object, build a query action, publish it as an MCP tool,

and connect it to Claude — which calls the tool live and returns real data. No

servers, no code.

Most tools that give agents database access (Supabase's MCP server, for example)

stick to raw SQL — this is meant for the layer above that.

It's live at [calamo.dev](https://calamo.dev). Mostly curious if this is solving a real problem for people

building agent workflows, or if I'm missing something obvious — feedback (and skepticism) welcome.