RAGFlow

RAGFlow is a broad platform for turning mixed documents into searchable knowledge and agent workflows. It is strongest for teams that need fine control over retrieval, model providers, and deployment, particularly in enterprise settings. Smaller teams should weigh its tight free-plan allowances and the hardware needed for a self-hosted installation.

Overview

RAGFlow combines document ingestion, retrieval, model connections, and agent orchestration in one workspace. Its ingestion pipeline cleans and processes multiple formats into semantic representations, making it suited to knowledge applications rather than a basic chatbot. Search blends vector retrieval with BM25, custom scoring, and advanced re-ranking, so teams can tune how results are selected and ordered.

Deployment is available through a browser-accessible service or Docker-based self-hosting. The documented minimum for a self-hosted build is 4 CPU cores, 16 GB of RAM, and 50 GB of disk, a meaningful commitment for individuals or small organizations. RAGFlow tests ARM64 platforms but does not maintain ARM Docker images; Linux ARM64 and macOS ARM64 users must build an image themselves.

Key features

Ingestion and retrieval

Connectors cover Notion, Google Drive, S3, Microsoft Teams, REST API, and Sitemap. The multi-format pipeline prepares those sources for retrieval, while hybrid search, custom scoring, and re-ranking provide more control than vector search alone. Source citations help readers trace answers back to supporting material, which matters in compliance, legal, and other evidence-sensitive work.

Agents and models

Visual workflows combine RAG, tools, and MCP servers for agent orchestration. You can connect online, local, and OpenAI-compatible model providers across knowledge bases, chats, search, and agents, avoiding a single-provider commitment. External MCP servers work over Streamable HTTP or SSE; stdio servers need a gateway or another service that exposes a supported endpoint.

Integration and safeguards

HTTP and Python APIs support application integration, and an assistant can be embedded in a third-party webpage through an iframe. By default, RAGFlow rejects MCP URLs resolving to loopback, private, link-local, reserved, or other non-public addresses, reducing exposure to internal network targets. Its privacy roles distinguish data it controls for its own purposes from customer data it processes for cloud users.

Pricing

RAGFlow uses a freemium model. The Free plan costs nothing and includes 5 Apps, 1 team member, 0.1 GB of dataset storage, and 500 credits per month. It has no API key, so it is mainly for individual evaluation or very small prototypes.

Starter costs $29 a month and raises limits to 50 Apps, 5 team members, 5 GB of dataset storage, and 5,000 credits per month, with an API key included. It fits a small working team, though the storage and credit ceilings still constrain heavier document collections or usage. Pro costs $129 a month, adds unlimited Apps, 20 team members, 50 GB of storage, and 20,000 monthly credits, while retaining API access; it is the practical tier for larger internal deployments.

Enterprise uses custom pricing and adds BYOC deployment, on-premises deployment, dedicated support, and a custom SLA. Those options target organizations that need controlled infrastructure and formal support rather than a fixed self-service subscription.

Platforms

RAGFlow supports web, API, Linux, macOS, and self-hosted deployment. The ARM image caveat makes standard x86 hardware the less complicated self-hosting path.

Who it's for

It suits enterprises and teams in financial services, legal and compliance, manufacturing, and education that need document-grounded applications, configurable retrieval, and agent workflows. Developers benefit from the HTTP and Python APIs and multiple model-provider choices. It is a weaker fit for solo users seeking generous hosted limits or anyone without the resources to meet its self-hosting baseline.

Pros and cons

• Pros: Hybrid search, custom scoring, and re-ranking give retrieval teams meaningful control.
• Pros: Visual RAG-and-tool workflows, MCP connectivity, and broad model support cover complete agent pipelines.
• Pros: Notion, Drive, S3, Teams, REST, and Sitemap connectors reduce bespoke ingestion work.
• Cons: The Free plan allows only one team member, 0.1 GB of storage, 500 credits monthly, and no API key.
• Cons: Self-hosting starts at 4 CPU cores, 16 GB RAM, and 50 GB disk, with no maintained ARM Docker images.

Alternatives

Choose AnythingLLM instead if you want Android and Windows support alongside web, API, Linux, macOS, and self-hosting, with a Docker-based open-source option. Flowise is a better starting point for a visual builder with a free trial and a Free plan capped at 2 flows or assistants, 100 predictions monthly, and 5 MB of storage. Pick PrivateGPT for a free, open-source deployment across web, API, desktop, and self-hosted platforms. txtai is the focused free Apache 2.0 package for pip or Docker installations. I3K RAG Enterprise offers a free AGPL-3.0 Community tier with Linux, ARM64, or Windows self-hosting. RAGStack is a free open-source package built around Astra DB Serverless or Apache Cassandra as its vector store. Dify provides a free Sandbox workspace with 200 message credits, one member, five apps, 50 knowledge documents, and 50 MB of storage. Haystack Enterprise Platform is worth considering for its free Studio workspace with one user, 100 pipeline hours, 50 files, and two development pipelines.

For a wider comparison, see the Retrieval-Augmented Generation Tools list.

Verdict

Choose RAGFlow when you need one platform for serious document ingestion, tunable hybrid retrieval, multiple model providers, and visual agent workflows, with cloud and controlled deployment paths. Look elsewhere if you need a generous free service, lightweight self-hosting, or ready-made ARM images; its limits and infrastructure demands are substantial before the higher tiers become worthwhile.