AutoRAG logo

AutoRAG

Free

Open source AutoML tool for RAG. Optimize the RAG answer quality automatically. From generation evaluation datset to deploying optimized RAG pipeline.

FreeFree tier
Type
Open Source
Company
Marker-Inc-Korea

About AutoRAG

AutoRAG is a self-evolving librarian agent that transforms document search from raw file listings to curated, actionable knowledge. It uses a two-tier agent workflow where a parent orchestrator delegates exploration to specialized explorer agents, searching PDFs, wikis, notes, research papers, and knowledge bases. Results are returned as clean, numbered knowledge units with page references, eliminating the need to manually open and read multiple files. A built-in self-evolving memory system learns from usage patterns and feedback, improving retrieval strategies over time. AutoRAG is fully open-source and free to use, designed as a customized Pi agent configured for document retrieval.

Key Features

Two-tier agent workflow: orchestrator delegates exploration to explorer agents
Self-evolving memory system that learns from queries and feedback
Returns curated, numbered knowledge units with page references
Supports multiple retrieval methods adaptable to different document types
Searches PDFs, wikis, notes, research papers, and knowledge bases
No raw file dumps — provides synthesized answers

Pros & Cons

Pros
  • Delivers curated answers with citations instead of raw file matches
  • Learns and adapts retrieval strategies over time based on usage
  • Handles multiple document types with a single interface
  • Fully open-source and free to use

Best For

Extracting key findings from quarterly reports and business documentsSearching and synthesizing information from research papers and academic literatureBuilding intelligent knowledge base querying for internal wikis and notesAutomating document review and summarization for legal or compliance teamsPersonal knowledge management across distributed digital notes