Dify is a visual IDE for AI: connect models, prompts, tools, and data through a drag-and-drop interface, then ship as a chatbot, REST API, or automated workflow. It supports 100+ LLMs — GPT-4o, Claude, Gemini, Llama, DeepSeek, and any OpenAI-compatible endpoint.

Free Tier: Cloud vs Self-Hosted

The real value is self-hosting. Deploy on any Linux server — even a free Oracle Cloud VM — for the full platform at zero usage cost. You pay only for the LLM API calls you make.

Self-Host with Docker

Docker Compose is the fastest path. The stack (API server, worker, web frontend, PostgreSQL, Redis, Weaviate vector DB, Nginx) comes pre-configured:

# Clone the repository
git clone https://github.com/langgenius/dify.git
cd dify/docker

# Copy and edit environment variables (optional)
cp .env.example .env

# Start all services
docker compose up -d

Dify then starts at http://localhost (port 80).

Resource Requirements

Oracle Cloud's Always Free tier (2 ARM cores, 12 GB RAM since the June 2026 cut) still runs a production Dify instance at zero cost.

Core Features

• Chatbot builder — visual prompt editor with system prompts, conversation memory, and context windows. Deploy as an embeddable widget or shareable link.
• Workflow (agent) builder — multi-step pipelines: LLM calls, tool use, conditional logic, HTTP requests, code execution. Ideal for document processing and content generation.
• RAG knowledge base — upload PDFs, Notion pages, URLs, or text; Dify chunks and embeds automatically so your bot answers from your data.
• 100+ model providers — OpenAI, Anthropic, Google, Mistral, Groq, DeepSeek, Ollama (local), and any OpenAI-compatible endpoint.
• Agent tools — built-in web search, calculator, code interpreter, Wikipedia, DALL-E, plus custom tools from any OpenAPI/Swagger spec.
• API + webhook publishing — every app exposes a REST API automatically, no extra backend needed.

Build Your First Chatbot

A customer-support bot with a knowledge base, in under 10 minutes:

1. Connect an LLM. In Settings → Model Providers, add your API key. For free options use Groq or a local Ollama model.

2. Create a knowledge base. Upload docs/FAQs (Dify handles chunking and vector indexing), or import via API:

curl -X POST 'http://localhost/v1/datasets' 
  -H 'Authorization: Bearer {dataset_api_key}' 
  -H 'Content-Type: application/json' 
  -d '{"name": "Support Docs"}'

3. Create the app. Create App → Chatbot → Basic, then set the prompt and attach your knowledge base under Context:

You are a helpful customer support agent for Acme Inc.
Answer questions based on the provided context.
If you don't know the answer, say "I'll connect you with a human agent."
Be concise and friendly.

4. Publish and integrate. Click Publish → API Access for your key, then call it from any app:

import requests

url = "http://localhost/v1/chat-messages"
headers = {
    "Authorization": "Bearer app-your-api-key",
    "Content-Type": "application/json"
}
payload = {
    "inputs": {},
    "query": "How do I reset my password?",
    "response_mode": "blocking",
    "conversation_id": "",
    "user": "user-123"
}

response = requests.post(url, headers=headers, json=payload)
print(response.json()["answer"])

Workflow Example: Document Summarizer

A workflow that accepts a URL, fetches its content, and returns a structured summary — built visually, then called via API:

# Workflow nodes (configured visually in Dify):
# 1. Start node — input: {url: string}
# 2. HTTP Request node — GET {url}
# 3. LLM node — prompt:
#    "Summarize this article. Return JSON with:
#     - title: string
#     - summary: 2-3 sentences
#     - key_points: list of 3-5 bullets
#     - action_items: list (if any)
#     Article: {{http_response.body}}"
# 4. End node — output: {result: LLM_output}

# Call the workflow via API:
curl -X POST 'http://localhost/v1/workflows/run' 
  -H 'Authorization: Bearer app-your-key' 
  -H 'Content-Type: application/json' 
  -d '{
    "inputs": {"url": "https://example.com/article"},
    "response_mode": "blocking",
    "user": "user-123"
  }'

Trigger Dify from Chat with OpenClaw

Pair Dify with OpenClaw to run your workflows from natural-language commands via WhatsApp or Telegram — no frontend code. Configure a custom HTTP tool:

{
  "name": "summarize_url",
  "description": "Summarize any URL using Dify",
  "method": "POST",
  "url": "http://your-dify-server/v1/workflows/run",
  "headers": {
    "Authorization": "Bearer app-your-key"
  },
  "body": {
    "inputs": {"url": "{{url}}"},
    "response_mode": "blocking",
    "user": "openclaw"
  }
}

Message OpenClaw "summarize https://example.com/article" and it calls your Dify workflow, returning a structured summary.

Dify vs Alternatives

Dify wins on the combination: a polished no-code UI, proper RAG pipelines, a visual workflow builder, multi-model support, and full self-hosting — all in one free package.

Who Should Use Dify?

• Developers: prototype AI apps in hours, expose as API, integrate into existing products.
• Small teams: build internal tools (support bot, doc Q&A) without hiring ML engineers.
• Startups: ship AI features with zero infra cost on Oracle Cloud's free tier.
• Enterprises: self-host for full data control, compliance, and unlimited scale.
• Agencies: deliver client chatbots with a repeatable, no-code workflow.

Final Recommendation

Dify is the fastest path from "I want an AI chatbot" to a production deployment. The self-hosted version is completely free with no usage limits — you pay only for LLM API calls, and with free tiers from Groq, Gemini, or DeepSeek you can run a full AI app at zero cost. For most teams building internal tools, support bots, or content pipelines, the workflow builder alone replaces what would otherwise need LangChain, a custom API server, and a React frontend.

Get started: dify.ai | GitHub (80k+ stars) | Documentation