Skip to main content

​
What is AgentOS?

AgentOS is a production-ready platform that transforms your AI agents into deployable, scalable services. When you deploy an agent to AgentOS, the platform automatically:
  • Creates API endpoints - Your agent becomes accessible via REST API
  • Manages infrastructure - Handles databases, message queues, and async workers
  • Provides observability - Tracks metrics, costs, tokens, and execution history
  • Handles deployments - Blue-green deployments, rollbacks, and versioning
  • Scales your agents - Automatic scaling based on demand
You focus on building intelligent agents, AgentOS handles everything else.

​
Installing Upsonic Framework

Install the Upsonic Framework to build your agents: 
pip install upsonic
The Upsonic Framework provides the tools to create AI agents with tasks, tools, memory, and more.

​
Starting an Agent Project

Create a new agent project using the Upsonic CLI: 
upsonic init
This command creates:
  • upsonic_configs.json - Agent configuration (required)
  • main.py - Your agent code (required)
Everything else is optional. You can add additional files, modules, and dependencies as needed.

​
Building your Agent

​
Required Files

Every agent needs two files:
  1. upsonic_configs.json - Configuration file
  2. main.py (or any .py file) - Agent implementation

​
Adding Dependencies

Add Python packages to your agent: 
upsonic install <package-name>
This command adds the package to upsonic_configs.json under dependencies.

​
Example Agent Code

Here’s a simple agent using the Upsonic Framework: 
from upsonic import Agent, Task

# Define your agent
agent = Agent(
    role="Research Assistant",
    goal="Help users find and analyze information",
    backstory="You are an expert researcher with access to various tools"
)

# Create a task
task = Task(
    description="Research the latest trends in AI",
    expected_output="A summary of current AI trends",
    agent=agent
)

# Execute
result = agent.execute_task(task)
print(result)
For more examples and detailed documentation, visit the Upsonic Framework Documentation.

​
Deploy your Agent to the Platform

AgentOS supports two deployment methods: Git and ZIP upload.

​
Deployment Process

Step 1: Upload Agent Code
Upload your agent to AgentOS using one of two methods:
  • Git Provider - Deploy from GitHub, GitLab, or Azure Repos
  • ZIP Upload - Direct upload without Git integration
Step 2: Configure Deployment
Specify deployment configuration:
  • Environment Variables - Set custom env vars for your agent
  • LLM Connection - Select the AI model to use
  • Machine Specification - Choose CPU and RAM allocation
Step 3: Start Deployment
Click Deploy to start the deployment process. AgentOS will:
  • Pull your code from Git or extract the ZIP
  • Install dependencies from upsonic_configs.json
  • Create containers and allocate resources
  • Start your agent and assign an API endpoint
Agent Deployment Configuration

​
Deploy from Git Provider

Deploy directly from your repository:
  1. Push your agent code to your Git repository
  2. In AgentOS, go to New Agent Deployment
  3. Select Deploy from Git
  4. Choose your repository and branch
  5. Configure environment variables and LLM connection
  6. Click Deploy
For detailed Git setup, see the Git Connection guide.

​
Deploy from ZIP

Deploy without Git integration:
  1. Compress your agent project folder into a ZIP file
  2. In AgentOS, go to New Agent Deployment
  3. Select Deploy from ZIP
  4. Upload your ZIP file
  5. Configure deployment settings
  6. Click Deploy
For more details, see the Git Connection - ZIP Deployment section.

​
Next Steps

Agent Management

Learn how to manage, monitor, and scale your deployed agents

Metrics & Monitoring

Track costs, tokens, response times, and execution history