How to Create Your First AI Agent Skill with Refly — Step-by-Step Tutorial

Getting Started with Refly: Build Your First Agent Skill

In this tutorial, you'll learn how to create a custom AI agent skill using Refly's vibe workflow editor — no coding required. By the end, you'll have a working skill that can search the web and summarize results.

Step 1: Deploy Refly

The easiest way to get started is with Docker:

docker run -d --name refly -p 8080:8080 reflyai/refly-web:latest

Once running, open http://localhost:8080 in your browser to access the Refly dashboard.

Step 2: Create a New Skill

In the Refly dashboard, click "New Skill". You'll see the vibe workflow editor — a blank canvas with a palette of node types on the left:

• Input — what the agent receives (user query, system prompt)
• Think — reasoning steps, including model selection and temperature
• Tool — external actions (web search, file read, API call)
• Output — how the response is formatted and delivered

Step 3: Build a Web Research Skill

Drag and connect these nodes to build a research assistant:

Step 4: Test and Share

Click Run to test your skill in the built-in preview panel. Once satisfied, click Publish to make it available. Your skill can now be used from:

• Claude Code — via the Refly MCP plugin
• Slack bot — team members type /refly-research
• API — POST to your Refly instance's API endpoint

Step 5: Deploy to Production

For team use, deploy with the full Docker Compose setup including PostgreSQL and Redis:

version: '3'
services:
  refly:
    image: reflyai/refly-web:latest
    restart: unless-stopped
    ports:
      - 8080:8080
    volumes:
      - ./data/refly:/data

Refly automatically handles database migrations and MCP server registration on first start.