Skip to main content

​
What is AgentOS?

AgentOS is a comprehensive platform that enables you to build and deploy AI agents with complete observability and management capabilities. With AgentOS, you can:
  • Build and deploy agents with streamlined workflows
  • Monitor everything - track metrics, token costs, API calls, response times, and execution history
  • Transform agents into APIs - make your agents API-native and production-ready
  • Manage deployments - handle the entire deployment lifecycle from a single interface
AgentOS handles migrations, databases, async job queues, and observability infrastructure automatically, letting you focus on building intelligent agents.
AgentOS Platform Overview

​
What is the Agent Framework?

The framework is where you develop your agent, add functionality, and design its behavior. It is a Python-based library that includes several features to help you build production-grade agents. Key capabilities include:
  • Create a knowledge base to handle large amounts of information
  • Create safety policies to prevent hallucinations and unsafe answers
  • Design your own memory structure to personalize your agent
  • Use Unified OCR to read all kinds of PDFs and images
Just focus on your use case—Upsonic Agent Framework will handle the rest.

​
Let’s Create an Agent That Gives Information About Stocks

In this agent, the user provides a stock symbol and we ask the agent to analyze it. We’ll use a specific response format to make the output useful and consistent.

​
Create an Agent Repository

1

1) Pick your GitHub organization

In GitHub, click New (top-right) or go to your organization page and click Repositories → New.In the Owner dropdown, make sure you select your Organization (not your personal account).
2

2) Fill repo details (Name / Visibility)

Choose a repository name (e.g. stock-agent).Select Visibility:
  • Private: Only accessible within your org (recommended).
  • Public: Visible to everyone.
Optionally add a short Description.
3

3) Add starter files (README is required)

Must do: check Add a README file so GitHub creates the repository with an initial commit.Optionally:
  • .gitignore: choose Python.
  • License: choose the license your org uses.
4

4) Create the repo and clone locally

Click Create repository.Then on the repo page: Code → copy SSH or HTTPS URL and clone locally:git clone <repo-url>

​
Install Upsonic CLI

We’ll use uv to manage the virtual environment and install the Upsonic CLI.
  1. Create a virtual environment:
uv venv
  1. Install Upsonic CLI into the environment:
uv pip install upsonic
  1. Verify the CLI works by printing the help output:
uv run upsonic --help
Screenshot2025 12 18at4 53 50PM

​
Create Agent Project With “upsonic init”

  1. Initialize a new agent project:
uv run upsonic init
  1. When prompted, type your agent name into the input (e.g. stock-agent) and press Enter.
Screenshot2025 12 18at5 17 44PM
  1. Install the project dependencies (required before running locally):
uv run upsonic install

​
Code your Inputs and Outputs

Now define what your agent accepts as input and what it returns as output by editing upsonic_configs.json.
  1. Open upsonic_configs.json and update the basic metadata:
    • agent_name: Your agent’s display name
    • description: What the agent does (one sentence is enough)
    • entrypoints.api_file: The python file that contains your agent entrypoint (e.g. main.py)
  2. Define your input(s) under input_schema.inputs:
    • Add one key per input field (example: user_query)
    • Set type, description, and required
    • type options: string, integer, dict
  3. Define your output(s) under output_schema:
    • Add one key per output field your agent returns (example: bot_response)
    • Set type and description
    • type options: string, integer, dict
Here is the relevant part of the config you will edit: 
{
  "agent_name": "Safety Engine Feedback Loop",
  "description": "Upsonic AI Agent that uses the Safety Engine Feedback Loop to ensure the output is compliant with the policy",
  "entrypoints": {
    "api_file": "main.py"
  },
  "input_schema": {
    "inputs": {
      "user_query": {
        "type": "string",
        "description": "User's input question for the agent",
        "required": true,
        "default": null
      }
    }
  },
  "output_schema": {
    "bot_response": {
      "type": "string",
      "description": "Agent's generated response"
    }
  }
}

​
Test in your local with “upsonic run”

  1. Run the agent locally:
uv run upsonic run
And then go to http://localhost:8000/docs . Screenshot2025 12 18at5 19 25PM Screenshot2025 12 18at5 20 11PM

​
Deploy to Upsonic AgentOS

1

1) Start a new deployment

From the main dashboard, click New Agent Deployment.Screenshot2025 12 18at5 22 15PM
2

2) Choose repository source

Click Deploy Existing Repositories.Screenshot2025 12 18at5 23 36PM
3

3) Select your agent repository

Find and click the agent repository you created (the repo you want to deploy).
4

4) Configure LLM Connection

Set up your LLM Connection (provider + credentials) for this deployment.Screenshot2025 12 18at5 24 47PM
5

5) Deploy

Scroll to the bottom and click Deploy.

​
Use your Agent

1

1) Open the deployed agent

Go to your agent from the AgentOS dashboard / agents list.
2

2) Find the API URL

In the agent menu, locate the API URL and click it.
3

3) Use the endpoint + Swagger docs

You will see an API endpoint page that includes the Swagger API documentation.