Skip to main content

​
What You’ll Build

A production-ready multi-agent system with three agents:
AgentWhat it does
PalAI-powered second brain. Captures notes, bookmarks, people, meetings, and research.
Knowledge AgentAnswers questions from a knowledge base (Agentic RAG).
MCP AgentConnects to external services via Model Context Protocol.
You’ll run it locally with Docker, then deploy to Railway.

​
Prerequisites

​
Run Locally

1

Clone the repository

git clone \
    https://github.com/agno-agi/agentos-railway-template.git \
    agentos-railway
cd agentos-railway
2

Configure environment

cp example.env .env
Add your OPENAI_API_KEY to .env
3

Start the system

docker compose up -d --build
4

Connect to the control plane

  1. Open os.agno.com
  2. Click Add OS → Local
  3. Enter http://localhost:8000
Your multi-agent system is now running:

​
Try Pal

Pal is your AI-powered second brain. It captures, organizes, and retrieves your personal knowledge. Capture: 
Note: decided to use Postgres for the new project - better JSON support
Bookmark https://www.ashpreetbedi.com/articles/lm-technical-design - great intro
Research: 
Research event sourcing patterns and save the key findings
Retrieve: 
What notes do I have?
What do I know about event sourcing?
Pal stores your data in DuckDB and uses the learning system to remember schemas, research findings, and error patterns.
Pal is a work in progress. It’s a starting point for building your own personal agent.

​
Deploy to Production

1

Install Railway CLI

brew install railway
2

Login and deploy

railway login
./scripts/railway_up.sh
The script provisions PostgreSQL, configures environment variables, and deploys your system.
3

Connect to the control plane

  1. Open os.agno.com
  2. Click Add OS → Live
  3. Enter your Railway domain
Your multi-agent system is now live.

​
Manage Your Deployment

railway logs --service agent_os      # View logs
railway open                         # Open dashboard
railway up --service agent_os -d     # Deploy updates

​
What’s Next

You now have a production multi-agent system with:
  • Learning - Agents that remember and improve over time
  • Persistence - PostgreSQL for agent state, DuckDB for user data
  • Research - Web search, company lookup, and people search via Exa
  • Monitoring - Full visibility via the AgentOS control plane

Add Your Own Agent

Extend the system with custom agents

Add Tools

Connect to Slack, Gmail, Calendar, and 100+ integrations

AgentOS Deep Dive

Learn more about the production runtime

Learning System

Configure how agents learn and remember