</>MCP Agents Market
Agent

AutoRAG

by Marker-Inc-Korea5.1kTypeScriptUpdated 2026-08-27

AutoRAG: Now your agent can find anything in your computer. It gets smarter if you are using it frequently.

AutoRAG is a self-evolving librarian AI agent that searches across local documents, PDFs, wikis, chat archives, and cloud drives to deliver curated, numbered answers instead of raw file paths. Built on the Pi agent framework, it federates data in place without forced migration, supports BM25, semantic vector, and hybrid retrieval through MinSync, and learns from user feedback to improve search strategies over time. The agent reads source files directly, judges evidence quality, and structures findings into actionable knowledge units. It integrates datasources like KakaoTalk, WhatsApp, Telegram, Slack, Discord, Notion, GitHub, Gmail, Google Drive, OneDrive, and Obsidian through external CLI tools, maintaining zero-copy federation with scope-checked access.

Key Features

Self-evolving memory system that learns optimal retrieval methods and productive document areas from usage patterns and explicit feedback
Federated search across local files, PDFs, and 15+ datasources (chat apps, email, cloud drives, Obsidian) without data migration
Multiple retrieval methods (BM25, semantic vector, hybrid) through MinSync with local EmbeddingGemma embedder requiring no API keys
Direct source reading via bash tools; agent curates structured SearchDocumentsResponse with numbered findings, not raw grep output
Jikji find-first discovery layer provides bounded answer packs while maintaining original file verification
Automatic duplicate detection via dupey CLI with exact-hash exclusion during indexing; read-only reporting never deletes files
Thin PDF extraction retry with docling-fast hybrid backend for sparse multi-page documents; automatic quality gating
P2P peer query sharing over SimpleX with injection screening, sharing policy enforcement, and manual approval workflow

Use Cases

  • 01Searching across internal documentation, manuals, research papers, and meeting notes with context-aware synthesis
  • 02Retrieving compliance requirements, legal citations, or technical specifications from large PDF collections
  • 03Unified search across Slack threads, Notion pages, Gmail archives, and Google Drive without separate tools
  • 04Building institutional knowledge retrieval for teams with chat history from KakaoTalk, Discord, WhatsApp, or Telegram
  • 05Obsidian vault semantic search with incremental qmd indexing and BM25/vector hybrid retrieval
  • 06Sharing curated research findings with trusted peers via P2P query server with PII redaction and approval gates

Related Agents

View more

hermes-agent

The agent that grows with you

Agent250.4k

agency-agents

A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.

Agent155.5k

openinterpreter

A coding agent for open models like Kimi K3

Agent68k

cline

Autonomous coding agent as an SDK, IDE extension, or CLI assistant.

Agent66.3k

AutoRAG — FAQ

What is AutoRAG and how does it differ from search tools?+

AutoRAG is a self-evolving librarian AI agent that searches, reads, judges evidence, and curates numbered knowledge units instead of returning raw file paths. It learns optimal retrieval strategies from usage and explicit feedback, federates datasources in place without migration, and structures answers into actionable findings.

How do I install AutoRAG?+

Install globally via Bun with 'bun install -g @autorag/librarian', then run 'autorag init' to create the config. Requires Node ≥24 or Bun, Java 11+ for PDF parsing, and optionally Rust/Cargo for MinSync and Jikji auto-install. External datasource CLIs (katok, discrawl, wacrawl, slacrawl, notcrawl, mailcrawl, rclone) install separately.

Which AI clients or runtimes does AutoRAG work with?+

AutoRAG resolves model and provider from the user's authenticated local runtime at search time. It does not ship a private provider default; you configure the model explicitly in config.json or via environment when no local authenticated provider is available.

Does AutoRAG require API keys or cloud services?+

No for basic document search. The local EmbeddingGemma embedder via Ollama requires no API keys and auto-installs. Datasource connectors like Gmail, Google Drive, Slack, Discord, Notion, or GitHub require their own tokens stored as environment variable references in config, but the agent uses local BM25/MinSync/Jikji by default.

Is AutoRAG free and open source?+

Yes, AutoRAG 2.0 is MIT licensed. The legacy Python AutoRAG (RAG AutoML pipeline optimizer) in the legacy/ directory is Apache-2.0 and continues in maintenance mode.

What datasources can AutoRAG search?+

AutoRAG federates local files, PDFs, and 15+ external datasources including KakaoTalk, WhatsApp, Telegram, Slack, Discord, Notion, GitHub issues/PRs, Gmail/IMAP via mailcrawl, Google Drive, OneDrive, iCloud (experimental), Obsidian vaults via qmd, RSS feeds, macOS Spotlight, and local screenshot/photo galleries via ClawGallery. Each uses its own CLI tool and maintains incremental archives in place.

How do I install AutoRAG?+

Open the source repository on GitHub and follow its README. AutoRAG is a agent — MCP Agents Market links you directly to the official repo.

Is AutoRAG free?+

AutoRAG is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches

Search AgentsKnowledge & Memory AgentsProductivity & Workflow AgentsBest MCP ServersBest SkillsBest PluginsTop rated listingsClaude Code toolsBrowse all categories