Project Description
zAira is an AI-powered data analyst built on top of Directus + MCP, enabling users to ask business questions in plain language and instantly get:
-
Auto-generated SQL queries
-
Meaningful data insights
-
Interactive visualizations (charts, summaries, trends)
-
Actions that write back to Directus via MCP
It turns Directus into an intelligent analytics platform—no coding, no SQL, no BI tools. Just chat.
zAira acts like a “ChatGPT for your Directus data,” but with the safety, structure, and governance of Directus collections.
Challenge Entered
Challenge 1: Build a project with Directus using MCP
zAira uses multiple MCP-powered actions, including:
-
MCP: Query Execution
AI generates SQL → MCP securely executes it against Directus -
MCP: Directus Schema Introspection
zAira fetches schema, relationships, field metadata to write proper SQL -
MCP: Visualization Integration
Generates ChartJS-ready datasets via a custom MCP tool
Technical Details
Architecture
-
Directus as the core data engine + collection store
-
Node.js MCP Server handling:
-
Schema introspection
-
Query construction
-
Query execution
-
Visualization formatting
-
-
LLM Agent using:
-
Schema cleanup helper
-
Structured Output for SQL + ChartJS formatting
-
MCP action calling
-
-
Front-end (optional)
-
Minimal chat interface or Directus extension
-
Visual rendering using ChartJS
-
Starting Point Details
We started the project using:
-
Directus Base template
-
Our existing LM Studio agent framework
-
A clean Directus instance + demo data (Inventory, HR, Sales mock data)
-
Baseline MCP action bindings for:
-
Schema
-
SQL
-
Items
-
-
Extended schema formatting functions and MCP tool wrappers (custom-built)
Initial goal:
Make analytics accessible for non-technical Directus users.
Final deliverable:
A plug-and-play AI analytics assistant for any Directus project.
Demo Video
