What is Dify? Open-source platform for building AI applications, workflows, RAG, and agents

What is Dify?

Dify is an open-source platform for building AI applications: you build chatbots, automated workflows, agents, and RAG systems in a workspace with a drag-and-drop interface, then publish as a web app or call via API. The name Dify, according to official documentation, is short for "Do It For You". Dify has a cloud version (Dify Cloud, with a free Sandbox plan and Professional plan at $590 per year) and a Community self-hosted version using Docker Compose; according to GitHub (read on 24/10/2026), the minimum configuration to run is 2 CPU cores and 4 GiB RAM.

This article goes through: 1) main features; 2) plans and pricing, checked on 24/10/2026; 3) how to self-host; 4) when to choose Dify over n8n or writing your own code; 5) OPCAIOS's real experience running Dify as an assistant on a website. Finally, there is a notes section. For example, a small company can build a chatbot that answers using internal documents without writing much code.

Dify brings workflow, RAG, agents, and model management into one place, published via web app or API. Diagram: OPCAIOS.

What is Dify?

Dify is an "open-source platform for building AI applications", according to Dify's official introduction page, including agents, agentic workflows, and chatbots using your own data. You publish as a web app or connect to your product via API. First, Dify brings everything into one place; next, every feature has an API. According to GitHub (read on 24/10/2026), the project has around 159.4 thousand stars and 25.2 thousand forks at the time of reading. The license is "Dify Open Source License", described as based on Apache 2.0 with additional conditions; please read the additional conditions carefully if you plan to use Dify to provide services to multiple customers. For example: an agency using one Dify instance to serve 20 customers may relate to the additional conditions, so you need to read the full text, which OPCAIOS has not yet done. If you're new to language models, see first What is an LLM?. There are three things to remember. First, Dify has a Cloud version and a Community version. Next, the license has additional conditions. Finally, star and fork counts change over time so they are only for reference.

What are Dify's main features?

Dify is a platform with seven groups of features, according to GitHub (read on 24/10/2026):

• Workflow: visual canvas to build and test AI processes.
• Model support: connect multiple LLMs from multiple providers, including self-hosted models and OpenAI-compatible API models.
• Prompt IDE: write prompts, compare performance between models, add features like text-to-speech.
• Pipeline RAG: from document ingestion to retrieval, extract text from PDFs, PPTs, and other formats.
• Agent: agents with their own sandbox, using skills, tools, MCP servers, or custom APIs.
• LLMOps: track logs, performance, improve prompts and data over time.
• Backend-as-a-Service: every feature has an API to integrate into other applications.

The documentation also mentions the difyctl command-line interface to run Dify applications from the terminal, scripts, CI, or other agents, plus a plugin system to extend tools, models, and integrations, and a Marketplace of community application templates. For example, a team can ingest documents into a Pipeline RAG, connect an OpenAI or Anthropic model, then call the application via API.

Dify Cloud: plans and pricing

According to the official pricing page (prices in USD, excluding tax; annual payment saves 17%):

	Sandbox	Professional	Team
Price	Free	$590 per workspace per year	$1,590 per workspace per year
Message credits	200 (cycle not stated)	5,000 per month	10,000 per month
Members	1	3	50
Number of apps	5	50	200
Knowledge bases	50	500	1,000
Knowledge storage	50 MB	5 GB	20 GB
Log history	30 days	Unlimited	Unlimited

Message credits are deducted by model type, and when exhausted you can switch to using your own API keys (listed providers include OpenAI, Anthropic, Gemini, xAI, Tongyi). The Community version is free, open-source, one workspace, you deploy yourself using Docker. The Enterprise version is priced by agreement, adds dedicated deployment, multiple workspaces, SSO, advanced security, SLA, and official support. For example, a team of 3 people using the Professional plan is just enough for the number of members. Pricing may change, please check the current pricing page.

How to Self-Host Dify with Docker?

Self-hosting Dify with Docker is how to run the Community edition on your own server, and according to Dify's GitHub (as of October 24, 2026), it involves four steps.

• Step 1: prepare a machine with at least 2 CPU cores, 4 GiB RAM, Docker and Docker Compose version 2.24.0 or higher (see What is Docker? and What is VPS?).
• Step 2: go to the docker directory of the source code, copy the sample environment file to .env.
• Step 3: start the containers in background mode.
• Step 4: open the /install path on your machine's address to initialize the admin account.

After that, put Dify behind a proxy server with HTTPS (see What is Nginx?), change all default passwords and keys in the environment file, and do not expose the internal ports of the database to the Internet. For example, a 2-core CPU, 4 GiB RAM VPS meets the minimum requirements for trying it out. Finally, back up the database before each Dify version upgrade.

When should you choose Dify?

You should choose Dify when you need to build an AI application with knowledge management and prompts, according to OPCAIOS's general assessment (not a measurement), across four situations.

• Chatbot or assistant using private documents, needing a knowledge management interface: choose Dify (integrated RAG).
• Multi-step workflow with conditional branches calling LLM, wanting to build quickly without writing much code: choose Dify workflow.
• Automating connections between hundreds of SaaS applications using events (webhook, schedule): choose tools like n8n or Make.
• Need control over every line of code, specific performance requirements: write code yourself with the model provider's SDK.

These two groups complement each other: you can use automation tools as an event layer and call Dify when you need an AI step. See how to connect via API in What is API? and What is Webhook?; for knowledge part see What is RAG?.

OPCAIOS's experience

OPCAIOS self-hosts a Dify instance on a private server and uses it as a "Personal Assistant" on the website. The assistant is a Chatflow with four steps in order: 1) classify customer needs; 2) query knowledge; 3) extract contact information; 4) transfer information to CRM. Instead of Dify's built-in chat frame, OPCAIOS built a custom interface on the website through a middleware layer on the server side that calls Dify using the app's publicly shared code (not embedding secret keys), with a limit of eight messages per minute for each IP address, and displays the status of each step ("Querying knowledge…") to the customer. The Dify service also sits in the list of systems checked by periodic monitoring, with alerts when it goes down. We don't have official metrics on the rate of customers leaving information through the assistant, so we're not including effectiveness numbers in the post.

Finally, the practical lesson: building quickly with drag-and-drop interface is very convenient, but user experience, abuse prevention and application code security still need to be handled by yourself at the middleware layer.

Notes when using Dify

• License: read Dify Open Source License's supplementary terms before commercial use.
• Data: the cloud version sends data through your chosen model provider; sensitive documents should consider self-hosting and internal models.
• Model cost: Dify does not cover the token costs of model providers; keep track.
• Abuse prevention: public applications need rate limiting and input validation.
• Updates: monitor releases and back up the database before upgrading.

Read more: How to create a self-learning AI Chatbot for internal documents (RAG) with Dify and Supabase.

Frequently Asked Questions

What is Dify?

It's an open source platform to build chatbots, workflows, agents and RAG systems using a visual interface, published via web app or API.

Is Dify free?

Dify has two free paths: Dify Cloud's Sandbox plan and the self-hosted Community edition. According to Dify's official pricing page (read on 24/10/2026), the Sandbox plan is limited to 200 message credits, 1 member and 5 apps. The Professional plan costs $590 per year per workspace, including 5,000 credits per month, 3 members and 50 apps. The Team plan costs $1,590 per year, including 10,000 credits per month, 50 members and 200 apps. Prices exclude tax; annual payment saves 17%. The Community edition is free and open source, but you cover server costs and maintenance yourself. For example, a single person might try Sandbox, while a 3-member team would need Professional. Enterprise plans are priced by agreement, and prices may change so check the current pricing page.

What's the minimum setup to self-host Dify?

According to the GitHub repository: 2 CPU cores, 4 GiB RAM, Docker and Docker Compose 2.24.0 or higher.

How is Dify different from n8n?

Dify focuses on AI applications (LLM, RAG, agents) with knowledge management and prompt control; n8n excels at automating workflows connecting multiple apps by events. You can use them together: n8n handles the event layer and calls Dify when you need an AI step.

Can I embed Dify on my own website?

Yes, via web app or REST API. Want to learn how to build an AI assistant and automate processes for your business? See OPCAIOS's AI application training course.

Source: docs.dify.ai (introduction), langgenius/dify GitHub repository and dify.ai/pricing page, read directly on 24/10/2026. Star counts, prices and limits change over time, prices exclude tax. We have not read the full text of Dify Open Source License. The "when to choose" comparison table is our general assessment. The experience is OPCAIOS's.