Kaelio/ktx provides an executable context layer for AI agents, enabling them to query data accurately through MCP with specialized skills, memory, and a semantic layer. It aims to improve AI agents' data analysis capabilities, especially for tools like Claude Code and Codex, by giving them better access to and understanding of complex data.
A starter prompt for Claude Code, what you'll need, and how to reach them.
You are Kaelio, the creator of ktx, a context layer for data and analytics agents. Your goal is to extend ktx to support more advanced semantic layer capabilities and a wider range of data sources, specifically focusing on integrating with a Neo4j knowledge graph for richer context. You need to provide a modular architecture that can be easily extended. The primary goal is to demonstrate how an AI agent can effectively query complex, interconnected data. Stack: Next.js 16 App Router, React 19, Tailwind v4, AI SDK v6 with Gemini, Neon Postgres (for metadata/agent states) and Neo4j (for the semantic layer/knowledge graph). MVP: Build a basic ktx-like framework that allows an agent to answer natural language questions by querying a simple, predefined Neo4j knowledge graph. This involves: 1) Setting up a Next.js project. 2) Implementing a Python backend service for Neo4j interaction and agent orchestration. 3) Defining a simple Neo4j schema (e.g., 'Company-HAS_PRODUCT-Product' relationships). 4) Creating an API endpoint in the backend that takes a natural language query, translates it (via LLM) into a Cypher query for Neo4j, executes it, and returns the result. 5) Building a simple React UI where a user can input a natural language query and see the agent's response based on the Neo4j data. Build/verify gate: A user can ask 'What products does Company X have?' and the agent successfully returns a list of products by querying the Neo4j graph.
Developers building MCP servers and AI tool integrations are the target audience; highlight how KTX enables richer, more accurate data interactions for their agents and tools.
ktx is an executable context layer for data and analytics agents 🐙 Allow Claude Code, Codex, and any AI agent to query data accurately through MCP with skills, memory and a semantic layer Topics: agent, agent-skills, agents, ai-agent, ai-agents, analytics, analytics-engineering, business-intelligence, claude, claude-code, claude-skills, codex, context-layer, data-analysis, data-engineering, llm, mcp, memory, semantic-layer, skills.
Open a pull request on the Kaelio/ktx GitHub repository or open a detailed issue outlining potential contributions.
“I've built a prototype integrating KTX-like functionality with a Neo4j knowledge graph, demonstrating how agents can answer complex queries from interconnected data. I'd love to contribute this to Kaelio/ktx to expand its semantic layer capabilities, or discuss how this could be a commercial offering to target developers struggling with data-grounded AI agents.”
Open the original ↗