What is RAGFlow?
RAGFlow is an open-source RAG engine and integrated agent platform that builds a context layer for AI agents by ingesting, structuring, and retrieving enterprise data. It pairs a built-in ingestion pipeline with hybrid retrieval and a visual workflow builder where RAG, tools, and MCPs are assembled in one place.
Key Features
- Built-in ingestion pipeline: RAGFlow cleanses and processes multi-format data — documents, images, and data sources — and structures it into rich semantic representations. This removes much of the manual preprocessing normally required before a dataset becomes retrievable.
- Hybrid retrieval with re-ranking: Retrieval combines vector search, BM25 full-text search, tensor-based matching, and custom scoring with advanced re-ranking, targeting higher answer accuracy and better context relevance.
- Dataset management: Datasets are created and managed inside the platform and feed directly into retrieval and agent workflows, keeping the retrieval corpus and the agents that consume it in the same system.
- Visual agent builder: Agents are assembled in an all-in-one visual workflow interface that connects retrieval, datasets, models, web search, chat, and MCPs without switching between separate products.
- MCP and HTTP tool integration: MCP servers and HTTP steps can be added as workflow nodes, so agents can call external services and internal systems alongside retrieved context.
- Multi-step agent orchestration: Workflows support subagents, iteration, conditional branching, and code steps, allowing agents to plan, branch, and reconcile internal records with external sources.
- Web search and model nodes: Agents can query web search and call configured models inside a workflow, mixing live external information with retrieved internal context.
- Enterprise deployment options: Enterprise plans add BYOC and on-premises deployment, dedicated support, and a custom SLA for production workloads.
Use Cases
- Financial research and stock analysis: A workflow identifies stock tickers from a user query, aggregates insights from external authoritative sources and internal records, combines them with financial metrics, and produces a complete investment report through multi-agent orchestration.
- Legal precedent analysis: A workflow extracts jurisdiction, court level, and legal issues from a case input, formulates search queries, retrieves relevant precedents with their ratio decidendi, and consolidates the results into a structured analysis with citations and distinguishing factors.
- Maintenance and technical support guidance: A workflow validates whether the input has enough detail, pulls standard protocols from internal maintenance manuals, supplements them with external technical references, and produces clear execution instructions.
- Enterprise knowledge assistants: Teams with large internal document sets can build a chat agent backed by re-ranked retrieval over managed datasets.
- RAG prototyping for engineering teams: Developers evaluating a RAG stack can use the free tier to test ingestion, retrieval, and agent workflows before committing to a paid plan.
Who It Is For
RAGFlow targets AI engineers, developer teams, and enterprise platform groups building retrieval-augmented agents over their own data. It addresses the work that usually sits outside the model itself: cleaning multi-format source material, building a retrievable dataset, improving retrieval accuracy with hybrid search and re-ranking, and wiring retrieval together with tools and models into a deployable agent workflow. Organizations with on-premises or BYOC requirements can run it under the enterprise plan.
Pricing
RAGFlow uses a tiered subscription model that starts with a free tier covering 5 apps, 1 team member, 0.1 GB of dataset storage, and 500 credits per month, without API key access. The Starter and Pro tiers are listed at $29/month and $129/month respectively, raising app counts, team members, dataset storage, and monthly credits and enabling API key access, while Enterprise plans add BYOC and on-premises deployment, dedicated support, and a custom SLA.
Frequently Asked Questions
What is RAGFlow used for?
RAGFlow is used to build a context layer for AI agents by ingesting and structuring multi-format data, retrieving relevant context through hybrid search, and assembling that retrieval with tools and MCPs into visual agent workflows. It is applied to tasks such as financial analysis, legal precedent research, and maintenance guidance.
Is RAGFlow free?
RAGFlow offers a free tier at $0 per month that includes 5 apps, 1 team member, 0.1 GB of dataset storage, and 500 credits per month, but no API key access. Paid Starter, Pro, and Enterprise tiers add higher limits, API key access, and enterprise deployment options.
Does RAGFlow support MCP and external tools?
Yes. RAGFlow's agent builder supports MCPs and HTTP steps as workflow nodes, so agents can call external services and internal systems alongside retrieval, datasets, models, and web search.
What retrieval methods does RAGFlow use?
RAGFlow combines vector search, BM25 full-text search, tensor-based matching, and custom scoring with advanced re-ranking. The platform presents these as complementary retrieval signals for improving answer accuracy and context relevance.
Is RAGFlow open source?
RAGFlow describes itself as an open-source RAG engine with an integrated agent platform, and publishes its source on GitHub alongside the hosted product. Enterprise-tier capabilities are offered separately through contact-based plans that include BYOC and on-premises deployment.