Context Layer for Data Agents 'ktx' Unveiled

ktx has been unveiled to solve the problem of general-purpose coding agents misunderstanding data warehouse schemas or generating incorrect metric calculation logic.

ktx automatically grasps the data stack to build a Semantic Layer, and acts as a context layer that consolidates business knowledge scattered across wikis and documents. Through this, instead of writing SQL from scratch every time, agents can declaratively call approved metrics to get accurate results.

Key Features and How It Works:

• Automatic Context Building: Automatically generates a wiki and semantic layer using databases (schemas, query history), BI tools (usage patterns), modeling code, and documents as sources.
• MCP (Model Context Protocol) Support: When various agents such as Claude Code and Cursor query ktx via MCP, ktx compiles the retrieved metrics into read-only SQL and executes it in the warehouse.
• Data Security and Local Operation: Operates locally without separate hosting, and warehouse connections are designed to be read-only, preventing the risk of database writes.
• Knowledge Integration: Beyond simple SQL generation, it combines documents (Wiki) and the semantic layer to help agents accurately understand business context.