RAGFlow
Open-source RAG engine with built-in ETL, hybrid search and visual agent orchestration for enterprise AI agents
RAGFlow is the pragmatic pick when your retrieval breaks on ugly PDFs and scanned tables, not on prompt design. The hosted $29/$129 promo tiers are cheap enough to trial before committing to self-hosting, though the free tier has no API key — fine for clicking around, useless for wiring into production. Pass if you want zero ops or a thin code library.
Verified 21h ago · liveness 88/100 · cite: rightaichoice.com/tools/ragflow
- Enterprises needing open-source RAG with BYOC or on-premises deployment for data control
- Teams building AI agents that need accurate retrieval from messy multi-format documents
- Financial services automating equity research into structured, cited reports
- Legal teams running structured precedent analysis across case law and internal matters
- Individual developers who want a lightweight code-only RAG framework — this is UI and deployment heavy
- Teams that want fully managed retrieval with zero infrastructure overhead
- Projects built on older GraphRAG or RAPTOR tutorials — both were deprecated in the v0.27 UI
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip RAGFlow if you want retrieval handled for you — running the Go backend means owning Docker hosts, the Elasticsearch vm.max_map_count setting, model provider keys and every upgrade yourself.
The Go rewrite changes your infrastructure bill: NATS replaces Redis as the message queue, so Redis-based monitoring and tuning you already have stops applying.
Pricing scales by apps, seats and dataset storage: Free gives 5 apps, 1 team member and 0.1 GB with no API key; Starter runs $29/mo promo then $59/mo for 50 apps, 5 members and 5 GB; Pro runs $129/mo promo then $259/mo for unlimited apps, 20 members and 50 GB. Enterprise is contact-sales for BYOC and on-premises. Against fully managed retrieval services you're paying infra yourself, but avoiding per-query markups on your own corpus.
In short
RAGFlow — Open-source RAG engine with built-in ETL, hybrid search and visual agent orchestration for enterprise AI agents. Best for Enterprises needing open-source RAG with BYOC or on-premises deployment for data control, Teams building AI agents that need accurate retrieval from messy multi-format documents, Financial services automating equity research into structured, cited reports. Free to start; paid plans from $29/mo.
What's new in RAGFlow
Checked 3 days agoAcross the latest 2 updates: 1 launch and 1 changelog entry.
RAGFlow v0.27.2: refactored Agentic RAG and new connectors
v0.27.2 refactors the Agentic RAG retrieval framework for faster reasoning. Adds knowledge-compilation rate limits, Sitemap and WebDAV sources, MonkeyOCRv2 and PaddleOCR-VL parsing, and Hubris and llmman model providers.
RAGFlow 0.27: Knowledge Compilation and Agentic Retrieval
Vendor post covering the 0.27 release, centered on knowledge compilation and an overhauled agentic retrieval pipeline.
What people actually say about RAGFlow — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
31 mentions across 3 sources (Hacker News, YouTube, GitHub) · researched Aug 24, 2026.
Average across the 3 sources that answered — each source counts once, not each post.
- +Hybrid search combining vector, BM25, and reranking delivers high retrieval accuracy.
- +Visual agent orchestration with MCP integration enables no-code workflow building.
- +Built-in ETL pipeline handles diverse data formats, from images to documents.
- +Self-hostable with full data control, appealing to privacy-focused enterprises.
- +Frequent updates including incremental connectors for Outlook, SharePoint, Slack, and more.
- −Unpatched security vulnerability raises serious deployment concerns.
- −Resource-intensive, needing robust hardware for smooth performance.
- −No ARM support limits deployment on inexpensive hardware.
- −UI-centric approach may deter code-first developers looking for lightweight frameworks.
- −Complexity of advanced features requires significant time to master.
- • Self-hosting requires significant hardware and maintenance expertise
- • Credits may run out quickly with heavy use
- • Enterprise pricing is opaque and likely high for full support
Viability Score
How well maintained and how widely used is RAGFlow? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Apache 2.0 open-source RAG engine, self-hostable
- Built-in ETL ingestion pipeline for documents, images and data sources
- Hybrid search combining vector search, BM25 full-text and custom scoring with reranking
- Layout-aware OCR parsing for complex tables and figures
- Parses PDF, DOC/DOCX, TXT, MD/MDX, CSV, XLS/XLSX, PPT/PPTX and images
- Knowledge compilation across Wiki, Graph, Tree, Page Index, Mind Map, Timeline and To Skills
- Agentic RAG with four thinking modes: Low, Medium, High and Ultra
- Visual AI agent orchestration integrating RAG, tools and MCP servers
- Go-based rewrite preview (v1.0.0-rc1) with NATS and Kvrocks
- MonkeyOCRv2 parsing and self-hosted PaddleOCR-VL deployment
- Sitemap and WebDAV data sources with incremental connectors
- Hubris and llmman model providers
- Knowledge-compilation rate limits to protect query throughput
- Deploy assistants to Discord, Feishu, WhatsApp, DingTalk and WeCom
- Retrieval API exposing rerank_candidates_count, knn top_k and num_candidates
About RAGFlow
RAGFlow is an Apache 2.0 RAG engine from InfiniFlow built as a context layer for AI agents rather than a chatbot wrapper. Teams pick it when the hard part isn't the model — it's getting clean, citable retrievals out of contracts, manuals, spreadsheets and screenshots. The platform handles ingestion through a built-in ETL pipeline with layout-aware OCR for messy tables and figures, then retrieves with hybrid search that blends vector, BM25 full-text and custom scoring plus reranking. Documents, images and connected data sources all feed the same dataset. Ingestion covers PDF, DOC/DOCX, TXT, MD/MDX, CSV, XLS/XLSX, PPT/PPTX and common image types. Recent releases added Sitemap and WebDAV sources alongside incremental connectors for Slack, SharePoint, Salesforce, OneDrive, Outlook, Microsoft Teams, Azure Blob and BigQuery. Knowledge compilation spans seven representations — Wiki, Graph, Tree, Page Index, Mind Map, Timeline and To Skills — and the v0.27.2 changelog applies rate limits to compilation so background jobs don't crowd out queries. Agentic RAG offers four thinking modes (Low, Medium, High, Ultra) to trade latency for reasoning depth per query. v0.27.2 refactored that retrieval framework for faster reasoning. A visual studio wires RAG, tools and MCP servers into multi-agent workflows, and finished assistants ship to Discord, Feishu, WhatsApp, DingTalk or WeCom. Parsing now includes MonkeyOCRv2 and self-hosted PaddleOCR-VL, with Hubris and llmman model providers added. Hosted plans start at $0 with a 5-App cap and no API key; Starter runs $29/mo (list $59), Pro $129/mo (list $259), and Enterprise covers BYOC, on-premises and custom SLA. Against managed retrieval services, RAGFlow asks you to own hosting but gives you the full pipeline to modify. Against code-only frameworks, you trade flexibility for a working UI.
Behind the Verdict
Start with the honest constraint: RAGFlow wants infrastructure. Whether you run the Go-rewrite preview (v1.0.0-rc1, NATS instead of Redis, Kvrocks for cache and checkpoints) or the stable line, someone owns upgrades, storage and scaling. That's the deal — you get data control and a pipeline you can actually modify. Where the tool earns its keep is document chaos. Layout-aware OCR, MonkeyOCRv2 and PaddleOCR-VL parsing matter when your corpus is financial statements with nested tables and legal exhibits with odd formatting. Hybrid vector plus BM25 plus reranking is the difference between plausible answers and quotable ones. We'd reach for it when retrieval quality is the measured bottleneck. The industry workflows on the vendor page — equity research, legal precedent analysis, manufacturing maintenance — are representative, not decorative: each needs cited output from internal and external sources in one pass. Knowledge compilation is the newer draw and the newer risk. Seven representations, from Graph to To Skills, give agents structured views beyond chunks. v0.27.2 added rate limits to compilation specifically because heavy jobs compete with queries, so budget compute accordingly. Agentic RAG's four thinking modes are genuinely useful for tuning cost per query. v0.27.2's refactor targets reasoning speed, which was the obvious complaint with high modes. Closest alternative depends on your pain. Want managed retrieval with nothing to babysit? A hosted vector or RAG service is the easier call. Want a lightweight library? LlamaIndex or LangChain will feel lighter. RAGFlow sits between: more opinionated than a framework, more yours than a SaaS. Caveats worth knowing. The free tier caps at 5 apps, one team member, 0.1 GB storage and 500 credits/month with no API key.
Researching RAGFlow? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas RAGFlow actually fits — and what changes day-one when you adopt it.
Upload broker reports, filings and internal notes into a dataset, run knowledge compilation to build a Graph and Tree over them, then wire a Search agent plus Report agent in the visual studio with You.com or Serply as a web fallback.
Outcome: A cited equity research report that merges financial metrics from filings with qualitative analyst commentary, refreshed incrementally rather than re-parsed from scratch each quarter.
Load public case law and internal matter records, then run the precedent-analysis pattern: a Formulate agent extracts jurisdiction, court level and legal issues, iteration retrieves similar cases, and a Report agent assembles the ratio decidendi with citations.
Outcome: Structured precedent memos with distinguishing factors ready for argument, generated from the firm's own corpus rather than re-researched manually per matter.
Deploy the Go backend on-premises, ingest internal maintenance manuals with SoMark OCR for the scanned ones, and set Agentic RAG to High thinking mode for troubleshooting queries via the WeCom or Teams assistant.
Outcome: Technicians get execution instructions sourced from the internal manual first, with external technical data used only as supplementary reference.
Use Cases
- Ingest scanned contracts with complex tables and build a RAG chatbot that cites exact clauses.
- Build a financial-report QA system that preserves table structure and numbers in answers.
- Self-host a privacy-sensitive RAG deployment for healthcare or legal documents.
- Generate a knowledge graph over a technical manual and answer questions from it.
- Automate equity research by aggregating financial metrics with qualitative insights via multi-agent planning.
- Analyze legal precedents by extracting jurisdiction, court level and legal issues, then comparing similar cases.
- Provide manufacturing maintenance guidance by extracting protocols from internal manuals with external references as backup.
- Deploy a customer support agent to Discord, Feishu, WhatsApp, DingTalk or WeCom using the visual agent workflow.
Models Under the Hood
as of 2026-10-16
Limitations
- Self-hosting is demanding: the quickstart recommends 4 x86-64 CPU cores, 16 GB RAM, 50 GB free disk, Docker 24.0.0 or newer with Compose v2.26.1+ and BuildKit, plus access to an LLM and embedding model at runtime.
- Linux ARM64 is not a supported native target for the Go image, and the default Elasticsearch document engine needs vm.max_map_count of at least 262144 on the host.
- Once you pick an embedding model for a dataset and parse a file with it, you cannot change it.
- The v1.0.0-rc1 preview's data upgrade from v0.27.2 runs automatically on container startup and is irreversible, deprecated APIs are no longer supported by the Go implementation, and known issues include DeepDoc supporting CPU environments only, the local sandbox no longer being supported, and the Team/Me permission model not yet being supported.
as of 2026-10-10
Verification history
We have re-verified RAGFlow 23 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 23 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published RAGFlow tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
A single person evaluating RAGFlow on a small sample corpus before committing to self-hosting.
What this tier adds
Starting tier: 5 apps, 1 team member, 0.1 GB dataset storage and 500 credits/month, with no API key.
Starter
$29/mo (promo), then $59/mo
Ideal for
A small team running a handful of internal RAG assistants against a modest document set.
What this tier adds
Adds API key access and raises limits to 50 apps, 5 team members, 5 GB dataset storage and 5,000 credits/month.
Pro
$129/mo (promo), then $259/mo
Ideal for
A growing department with production assistants, more contributors and a larger corpus to index.
What this tier adds
Adds unlimited apps, 20 team members, 50 GB dataset storage and 20,000 credits/month.
Enterprise
Custom
Ideal for
Organizations with residency, deployment or SLA requirements that rule out a shared hosted environment.
What this tier adds
Contact-sales tier adding BYOC and on-premises deployment, dedicated support and a custom SLA.
Where the pricing makes sense
The company stage and team size where RAGFlow's pricing actually pencils out — and where peers do it cheaper.
Pricing scales by apps, seats and dataset storage: Free gives 5 apps, 1 team member and 0.1 GB with no API key; Starter runs $29/mo promo then $59/mo for 50 apps, 5 members and 5 GB; Pro runs $129/mo promo then $259/mo for unlimited apps, 20 members and 50 GB. Enterprise is contact-sales for BYOC and on-premises. Against fully managed retrieval services you're paying infra yourself, but avoiding per-query markups on your own corpus.
Setup time & first value
How long it actually takes to get something useful out of RAGFlow — broken out by persona, not the marketing-page minute.
On a prepared x86-64 host with Docker and Compose already installed, the quickstart path — clone, check out the Go tag, docker compose up, verify /api/v1/system/healthz returns 200 — is roughly an hour. Add an LLM provider API key and create your first parsed dataset within that same session. Realistically budget a few days to a week to a first useful deployment: OCR on large scanned corpora is
Switching to or from RAGFlow
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From GraphRAG or RAPTOR pipelines: previously generated content stays searchable, but rebuild on the Graph and Tree knowledge compilation representations, which replaced both in the v0.27 UI.
- →From a code-only RAG library such as LangChain: move document ingestion into RAGFlow datasets and chunk templates, then re-express your retrieval chain as a visual agent workflow with MCP tools.
- →From v0.27.2 to v1.0.0-rc1: back up all data first, confirm you are on v0.27.2, then start the Go container — the migration is automatic and irreversible.
- →From a managed retrieval service: export vectors and documents, then re-parse into RAGFlow datasets, since each dataset is bound to the embedding model it was first parsed with.
- ↗To a managed retrieval service: export documents and embeddings from your datasets, accepting that layout-aware parsing and knowledge compilation output do not transfer.
- ↗To a code-only RAG framework: re-implement retrieval and reranking yourself; you lose the built-in ETL, chunk templates and the visual agent orchestrator.
- ↗To the v1.0.0 release once it lands: the Go rewrite is already the upgrade path, but deprecated APIs removed in the Go implementation must be replaced in your integrations first.
Integrations
Resources & Guides
- Documentationragflow.io
Quickstart
RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding. When integrated with LLMs, it is capable of providing truthful question-answering capabilities, backed by well-founded citations from various complex formatted data.
- Resourceragflow.io
Blog
Blog
- Resourceragflow.io
Tutorial - Build an E-Commerce Customer Support Agent Using RAGFlow
Currently, e-commerce retail platforms extensively use intelligent customer service systems to manage a wide range of user enquiries. However, traditional intelligent customer service often struggles to meet users’ increasingly complex and varied needs. For example, customers may
- Resourceragflow.io
Tutorial - Building a SQL Assistant Workflow
Workflow overview
- Resourceragflow.io
Changelog
Key features, improvements and bug fixes in the latest releases.
- Resourceragflow.io
RAGFlow in Practice - Building an Agent for Deep-Dive Analysis of Company Research Reports
In the actual work of the investment research department of financial institutions, analysts are exposed to a vast amount of industry and company analysis reports, third-party research data, and real-time market dynamics on a daily basis, with diverse and scattered information so
Tutorials & Learning
FREE: Ragflow Open-Source GUI AI Agent Rag Graph!🤖 Vector & Semantic Search Self-Hosted 🔍
Josh Pocock
YouTube returned 6 videos for “RAGFlow”, and we withheld 5: 5 could not be judged, because “RAGFlow” is a single word that other videos use for other things. Showing the 1 we can prove is about RAGFlow.
Official links
Tools that pair well with RAGFlow
Common stack mates teams adopt alongside RAGFlow, with the specific reason each pairing earns its keep.
OpenAgents
Open-source research platform (COLM 2024) shipping three language agents — Data, Plugins and Web — in one chat UI, for free.
MLflow
Open source AI engineering platform for agent and LLM tracing, evaluation, prompt management, and ML lifecycle work.
DataRobot
Enterprise agent workforce platform for building, operating, and governing AI agents across on-prem, hybrid, and multi-cloud environments.
Featured Head-to-Head Comparisons
Alternatives to RAGFlow
View all
OpenAgents
Open-source research platform (COLM 2024) shipping three language agents — Data, Plugins and Web — in one chat UI, for free.
Frequently Asked Questions
Categories
Best-of guides
Used RAGFlow? Help shape our editorial sentiment research.