Overview

XERJ revolutionizes data search for AI, acting as an intelligent auto-indexing engine that allows agents to understand your data without the token inefficiency of traditional tools like grep and sed. By indexing code, documentation, logs, and PDFs with a single command, XERJ drastically reduces token waste (up to 40x less than grep), making it an ideal solution for tasks such as RAG, security audits, and agent memory. Its Elasticsearch compatibility ensures seamless integration with existing clients and workflows.

Key Features

• Auto-Indexing: Automatically indexes any folder, making entire datasets accessible to AI agents without manual configuration.
• Token Efficiency: Achieves significant token savings (up to 40x less than grep) by indexing data intelligently, allowing agents to retrieve precise information rather than entire files.
• Versatile Data Support: Indexes a wide range of file types including code, documentation, logs, PDFs, CSVs, JSON, and more, even handling complex formats like Unity projects and emails.
• Elasticsearch Compatibility: Speaks the Elasticsearch API, allowing seamless use with existing clients, dashboards, and tooling.
• Agent Integration: Designed for AI agents, with features like MCP (Model Context Protocol) support for seamless integration into agent workflows and /_memory for durable agent memory.
• Local-First & Offline Embeddings: Provides a default lexical embedder that works offline, and an optional neural embedder that can be downloaded for advanced semantic search.
• Single Static Binary: Easy installation with a single binary, requiring no JVM or external dependencies.

Typical Use Cases

• Reference Coding: Enables AI agents to intelligently search and leverage existing open-source codebases similar to the current task, drastically improving code generation efficiency and reducing output tokens.
• Codebase Q&A and RAG: Allows agents to ask questions about a codebase and receive precise passages with file and line number context, significantly improving retrieval accuracy for RAG applications.
• Security Audits: Index target trees and query for vulnerabilities, secrets, and dangerous patterns across all file types simultaneously.
• Log and Incident Analysis: Transforms mixed-format logs into typed, queryable indices comparable to Elasticsearch, simplifying analysis and incident response.
• Agent Long-Term Memory: Provides a durable, recallable memory system for AI agents using the /_memory API, allowing them to retain and retrieve learned information across sessions.