Show HN: Nexus, ask AI about sensitive spreadsheets locally (github.com)

Nexus, a newly launched local-first agent layer for structured data, allows users to query and manipulate files like CSV, XLSX, and SQLite on their own machines without transferring any data to the cloud. Unlike existing AI tools that require uploading data—such as Salesforce with ChatGPT or Google Sheets with Gemini—Nexus keeps data local and leverages a Master Control Protocol (MCP) server to interact with AI agents like Claude Code and Cursor. This innovative approach ensures users maintain control over their sensitive information, which is a significant advancement in data privacy within the AI/ML community. The platform introduces several key features, including universal input support for various tabular data formats, non-destructive derivations for saving subsets and what-if scenarios, and an agent-native environment that allows AI tools to interact with structured data seamlessly. By implementing a local LLM named Iris, Nexus enhances user experience by auto-generating column types, summaries, and suggested queries. Upcoming updates will include selective cloud publishing options, allowing users to share specific data derivations while keeping the master dataset private. This development positions Nexus as a powerful solution for users who prioritize data security while exploring AI capabilities.