Open-Source RAG Platforms & Deep Research Agents: RAGFlow, MaxKB, GPT Researcher, DeerFlow & More

Answering questions from your own data and researching open-ended topics look similar on the surface, but they rely on different architectures. Document RAG platforms (like RAGFlow and MaxKB) focus on layout-aware PDF parsing, chunking, vector indexing, and grounded Q&A over a curated corpus.

Deep research agents (like GPT Researcher, ByteDance's DeerFlow, and MiroThinker) plan sub-questions, run dozens of live web and document searches in parallel, and synthesize multi-page cited reports—while code-graph tools like code-graph-rag parse source repositories with Tree-sitter into a graph database.

AgentIndex · Updated 2026-10-20

Knowledge and research tools by task

Document RAG indexes your files; research agents synthesize reports; graph RAG maps source code.

Document RAG platforms
PDF parsing, chunking and cited Q&A.
- ragflow
- MaxKB
Deep research agents
Multi-step search and long cited reports.
Codebase Graph RAG
Tree-sitter AST graph in Memgraph.
- code-graph-rag

Quick picks

ragflow — Apache-2.0 RAG engine with deep document parsing (layout analysis, OCR, tables), template chunking, and traceable citations.
MaxKB — GPL-3.0 all-in-one knowledge-base and workflow platform from the 1Panel team that starts with a single Docker command.
GPT Researcher — Apache-2.0 autonomous deep research agent that searches web and local files in parallel to write 2,000+ word cited reports.
deer-flow — ByteDance's MIT-licensed LangGraph super-agent runtime with a code sandbox, filesystem, memory, and Feishu/WeCom/Slack bots.
MiroThinker — MIT-licensed open research-agent models (30B/235B), MiroFlow framework, and 147k-sample training dataset for long-horizon tool use.
code-graph-rag — MIT-licensed graph RAG system that parses multi-language codebases with Tree-sitter into Memgraph and exposes an MCP server.
haystack — mature Apache-2.0 modular Python framework for custom RAG pipelines, reranking, and Hayhooks REST/MCP endpoints.

Side-by-side comparison

Tool	What it is	Stars	License	Last push	Maintenance	Releases, 90 days	Install
ragflow	Open-source RAG engine with deep document parsing and citations	92.6k	Apache-2.0	2026-10-21	Active	5	docker compose -f docker-compose.yml up -d
MaxKB	Open-source platform for knowledge-base agents and RAG	23.0k	GPL-3.0	2026-10-21	Active	4	docker run -d --name=maxkb --restart=always -p 8080:8080 -v ~/.maxkb:/opt/max…
GPT Researcher	Open-source deep research agent that writes cited reports	30.2k	Apache-2.0	2026-10-10	Active	2	pip install gpt-researcher
deer-flow	ByteDance's open-source super-agent runtime for long tasks	84.3k	MIT	2026-10-22	Active	1	uv sync
MiroThinker	Open-source deep research agent models and framework with long-horizon tool use	8.4k	MIT	2026-07-06	Slowing	0	git clone https://github.com/MiroMindAI/MiroThinker
code-graph-rag	Builds a graph of your codebase with Tree-sitter and lets you query it in natural language	5.3k	MIT	2026-10-22	Active	18	git clone https://github.com/vitali87/code-graph-rag.git
haystack	deepset's mature open-source Python framework for RAG, search and agents built from modular pipelines	26.8k	Apache-2.0	2026-10-21	Active	12	pip install haystack-ai
AgentGuide	Chinese study guide that links AI agent learning resources to a job-search path	11.1k	n/a	2026-09-15	Active	0	git clone https://github.com/adongwanai/AgentGuide.git
pipeshub-ai	Open-source workplace AI search with permission-aware answers	3.8k	Apache-2.0	2026-10-22	Active	7	curl -fsSL https://get.pipeshub.com/install \| bash

Tools in this list with an in-depth AgentIndex review: 9/9

How to choose

If your source material is complex PDFs, scanned documents, and spreadsheets, prioritize layout analysis and visual chunk inspection (RAGFlow).
If you need a fast-deploying internal Q&A system with domestic Chinese and global LLM support, check single-container platforms (MaxKB).
For multi-page cited web and document research reports, pick a dedicated research agent (GPT Researcher) or a sandboxed long-task harness (DeerFlow).
Check hardware requirements: RAGFlow and DeerFlow recommend 16 GB+ RAM, and MiroThinker's 30B/235B open models require dedicated GPUs.

FAQ

What is the difference between RAGFlow and MaxKB?

RAGFlow (Apache-2.0) goes deeper on complex PDF layout analysis, table extraction, and manual chunk inspection, at the cost of a heavier multi-service deployment (16 GB RAM recommended). MaxKB (GPL-3.0) deploys via a single Docker command and is geared towards rapid enterprise Q&A embedding and workflow orchestration.

Should I use GPT Researcher or DeerFlow for deep research?

Use GPT Researcher if you want a focused Python library or web app that turns a topic into a cited PDF/Word report. Use DeerFlow if you want a full agent runtime with a bash/Docker sandbox, persistent memory, sub-agents, and triggers from Feishu, WeCom, or Slack.

Should I pick Haystack for code-first RAG or RAGFlow for an out-of-the-box knowledge base?

Pick Haystack if you are a Python developer building tailored pipeline logic with custom rerankers, guardrails, and conditional branches. Pick RAGFlow if you want a turnkey self-hosted knowledge base with deep visual document chunking (OCR/tables) and a ready-made QA interface.

How we compiled this list

Tools are chosen by hand from the projects in our index, favouring those we have reviewed in depth. Stars, licence, last-push date and install command come from each repository and are refreshed with the site data. The maintenance column comes from the GitHub API (archived flag, last commit, releases in the last 90 days) and is refreshed daily. We read each project's README and docs; we have not run every tool ourselves, so test before you depend on one.