What is Dify?
Dify is a platform for building LLM applications. The vendor describes it as combining visual AI workflows, a RAG pipeline, agent capabilities, model management and observability. It is offered as hosted Dify Cloud, a self-hosted Community Edition and a self-hosted Enterprise edition.
Who it's for
The vendor describes it as a platform for production agentic workflows, for builders and enterprises moving from prototype to production. Enterprise is described as built for procurement, security and audit needs such as SSO, role-based access control and audit logs.
Pros & cons
Strengths
- Self-hosting is offered. The free Community Edition runs with Docker Compose (the README lists at least 2 CPU cores and 4 GiB of RAM), and Enterprise self-hosting is available through sales
- The README says it connects to hundreds of proprietary and open-source LLMs, including OpenAI API-compatible models
- The README lists a visual workflow canvas, a RAG pipeline with extraction from PDFs and slides, agents that can use Marketplace tools, MCP servers or custom APIs, and logs
- The README says every Dify offering comes with an API
- The free Sandbox plan has 1 workspace, 5 apps, 50 knowledge documents and 30 days of log history, and you can switch to your own API key when message credits run out
Limitations
- Sandbox allows 1 member, 200 message credits (the page does not say monthly), 5,000 API requests a month and 3,000 trigger events
- Cloud plans are per workspace with one workspace each. Professional allows 3 members and 50 apps, Team allows 50 members and 200 apps. Multiple workspaces are listed only under Enterprise
- Message credits are used per AI response at a rate that varies by model. Some models cannot be used with credits and need your own API key
- The licence is not plain Apache 2.0. It adds conditions on multi-tenant use and on keeping the logo and copyright in the console, and commercial authorization is required in some cases
- Some features are tier-gated or not yet available: LangSmith and Langfuse analytics are listed for paid plans, app version control is listed as coming soon and pay by invoice is listed as being explored
Plans and pricing
- Sandbox (Cloud): $0. 200 message credits, 1 member, 5 apps, 50 knowledge documents, 50 MB storage, 5,000 API requests a month, 30 days of logs.
- Professional (Cloud): $590 per workspace a year billed annually, or $59 a month billed monthly. 5,000 message credits a month, 3 members, 50 apps, 500 knowledge documents, 5 GB storage, unlimited log history, priority email support.
- Team (Cloud): $1,590 per workspace a year billed annually, or $159 a month billed monthly. 10,000 message credits a month, 50 members, 200 apps, 1,000 knowledge documents, 20 GB storage, priority workflow execution.
- Community (self-hosted): $0. Core features from the public repository, a single workspace, under the Dify Open Source License. No numeric limits are stated.
- Enterprise (self-hosted): Custom, contact sales. Commercial licence, multiple workspaces, SSO, security controls, official updates and support. No figures are stated.
Prices are from dify.ai/pricing and its Cloud page, read on 18 October 2026, and exclude taxes. The main page shows annual prices; the monthly amounts are on the Cloud page. The card's $49 a month for Professional is the $590 annual price divided by 12, which is our arithmetic, not the vendor's. Licence wording is from the repository README and LICENSE file: check the current text before relying on it for commercial or multi-tenant use.
Details checked October 2026 against the vendor's own pages. Prices and limits change, so confirm before you buy. Official pricing page ↗
Sources
This profile was compiled from these public pages. We have not run hands-on tests.
Dify — at a glance
Dify is a platform for building LLM applications. The vendor describes it as combining visual AI workflows, a RAG pipeline, agent capabilities, model management and observability. It is offered as hosted Dify Cloud, a self-hosted Community Edition and a self-hosted Enterprise edition.
Try Dify →