ktx AI Data Agents Context Skill

Skill by ara.so — MCP Skills collection.

ktx is a self-improving context layer that teaches AI agents how to query your data warehouse accurately. It combines approved metric definitions, joinable columns, wiki knowledge, and dbt/Looker metadata into one searchable surface for agents like Claude Code, Codex, Cursor, and OpenCode.

What ktx Does

• Auto-learns warehouse structure: Samples tables, detects joinable columns, resolves fan/chasm traps
• Ingests company knowledge: Combines dbt, MetricFlow, LookML, Looker, Metabase, and Notion content
• Builds semantic layer: Creates reusable metric definitions with automatic join resolution
• Serves agents via MCP: Exposes CLI and Model Context Protocol tools for agent execution
• Flags contradictions: Identifies conflicts across wiki pages and metric definitions
• Read-only by design: Never writes to your warehouse

Installation

Global Installation