Building a custom AI app used to mean hiring developers and waiting months. Dify changes that: it is an open-source platform for building AI applications — chatbots, agents, text generators, and RAG pipelines — through a visual interface, then deploying them with an API. This tutorial shows you how to install Dify, connect a language model, and ship your first AI app for your small business.
What It Is
Dify is an open-source LLM app development platform. Its visual interface combines AI workflows, RAG (retrieval-augmented generation) pipelines, agent capabilities, model management, and observability, letting a team go from prototype to production without rebuilding the stack. As checked on October 18, 2026, it has 158,748 stars on GitHub (github.com/langgenius/dify) and is licensed under the Dify Open Source License, which is based on Apache 2.0 with additional conditions.
In practical terms, Dify gives you five building blocks: Workflow (visual canvas for AI logic), model support (hundreds of proprietary and open-source models from dozens of providers, including GPT, Mistral, Llama 3, and any OpenAI API-compatible model), Prompt IDE (craft and compare prompts), RAG pipeline (ingest PDFs, PPTs, and documents so the AI answers from your files), and Agents (autonomous agents with tools, skills, and MCP servers). Everything ships with APIs, so you can embed Dify apps into your own website or business software.
Who It’s For / SMB Use Case
Dify fits small businesses that want a branded AI experience without building infrastructure. Practical uses include:
- Customer support chatbot: Train a chatbot on your FAQs, policies, and product docs so it answers common questions 24/7, with your branding and your tone.
- Internal knowledge assistant: Let staff query company handbooks, SOPs, and past project files in plain language instead of digging through shared drives.
- Content and document workflows: Build apps that summarize contracts, draft proposals from templates, or turn meeting notes into action items.
- Lead qualification agent: An agent that chats with website visitors, asks qualifying questions, and books qualified leads into your calendar.
Dify is not a no-brainer install for everyone: self-hosting needs a server with at least 2 CPU cores and 4 GB of RAM, plus Docker know-how. If you have no one technical on the team, the hosted Dify Cloud version is the simpler path — you trade some control for zero server maintenance.
Installation
The README requires minimum specs of CPU >= 2 cores and RAM >= 4 GB, plus Docker and Docker Compose v2.24.0 or later installed on your machine. The Docker Compose method is the easiest way to start the server.
Self-host with Docker Compose
Clone the repository and start the stack:
git clone https://github.com/langgenius/dify.git cd dify cd docker cp .env.example .env docker compose up -d
After the containers are running, open your browser at http://localhost/install and complete the initialization to create your admin account. The README points to the FAQ in the docs if setup fails.
Cloud alternative
If you don’t want to manage a server, Dify Cloud hosts everything for you — sign up at cloud.dify.ai and skip the commands above. Check the pricing page on dify.ai for current plans and usage allowances. For organizations needing SSO, RBAC, and support SLAs, there is a Dify Enterprise edition.
How to Use Dify
Step 1: Add a model provider
After logging in, go to Settings → Model Provider and add your API key for a provider such as OpenAI, Anthropic, Google, or a self-hosted model. Dify supports hundreds of models; pick one good general model (for example GPT or Claude class) for chat and one cheaper model for background tasks. Without a configured model, apps cannot generate text.
Step 2: Create your first app
Click Create App and choose a type: Chatbot (Q&A over your data), Agent (autonomous, uses tools), Text Generator (single-turn tasks like summarizing), or Workflow (multi-step logic on a canvas). For a support bot, choose Chatbot. Name it after its job, e.g. “Acme Support Assistant”.
Step 3: Feed it your documents (RAG)
In the app’s Knowledge section, upload your FAQs, policy PDFs, product manuals, or price lists. Dify extracts text from PDFs, PPTs, and common document formats, chunks it, and indexes it for retrieval. Good knowledge hygiene matters: keep documents current, remove duplicates, and write a short instruction telling the app how to answer (“Answer only from the uploaded documents; say you don’t know when unsure”).
Step 4: Design the prompt and test
Use the Prompt IDE to write the system instructions, set the tone, and compare how different models answer the same question. Test with real customer questions — including tricky ones — before anyone else sees the app. Iterate on the prompt until answers are accurate and on-brand.
Step 5: Publish and embed
Click Publish, then use the provided share link, embed code, or API to put the app on your website, in your help center, or inside your internal tools. Because every Dify app ships with an API, your developer can wire it into a CRM or custom dashboard. Monitor the LLMOps logs to see real conversations and keep improving prompts based on production data.
Pricing / Limitations
The self-hosted Community Edition is free under the Dify Open Source License (Apache 2.0-based) — you pay for the server and the model API usage, not for Dify itself. Note the license has additional conditions beyond plain Apache 2.0, so read it before commercial redistribution. Dify Cloud has paid plans (check the pricing page), and Dify Enterprise adds SSO, RBAC, and SLAs for organizations.
Limitations: self-hosting needs ongoing maintenance (updates, backups, monitoring). RAG quality depends on your documents — messy, outdated files produce messy answers, so budget time for knowledge curation. Model API costs scale with usage; a popular public chatbot can get expensive, so set usage limits and consider a cheaper model for high-volume traffic. As with any AI, review answers on sensitive topics and keep a human in the loop for anything irreversible.
Bottom Line for SMB
Dify is the fastest credible path from “we should have an AI chatbot” to a working, branded app trained on your own documents — in an afternoon, not a quarter. Start with one high-value use case like customer support or internal knowledge search, prove it works, then expand to agents and workflows. For prompt ideas to feed your new app, see our AI prompt templates, and for writing help, our best AI writing tools roundup.
Frequently Asked Questions
Do I need to code to use Dify?
No. Apps are built through the visual interface — you configure models, upload documents, and design prompts without writing code. Code is only needed if you want deep custom integrations via the API, which is optional.
Is my business data safe with Dify?
With self-hosting, your documents and conversations stay on your own server — Dify doesn’t see them. With Dify Cloud, your data is processed on Dify’s infrastructure under their terms, so check the privacy policy and consider self-hosting for sensitive data.
Which model should I connect first?
Start with a well-known general model from a major provider (OpenAI, Anthropic, or Google class) for the best out-of-the-box quality, then experiment with cheaper or open-source models once your prompts are stable. Dify lets you switch providers without rebuilding the app.
