AgentOS Features

Multi-LLM

Universal LLM Support

Connect to any major LLM provider with a single configuration change. Switch between OpenAI, Claude, Gemini, Cohere, GitHub Models, or run locally with Ollama - all without changing your code.

6+ providers supported out of the box
Simple provider switching with flags
Local models with Ollama (no API key needed)
Consistent API across all providers

GitHub

◯ OpenAI

A Claude

Gemini

C Cohere

Ollama

Chat Mode

Interactive Chat Interface

Engage in real-time conversations with any LLM through our beautiful terminal interface. Features markdown rendering, syntax highlighting, and persistent chat history.

Rich terminal UI with colors
Markdown rendering in responses
Chat history management
Customizable temperature and prompts

agentos chat --provider claude --temperature 0.7

You

Create a function to calculate fibonacci numbers

Here's an efficient fibonacci function using memoization:

def fibonacci(n, memo={}):
    if n in memo:
        return memo[n]
    if n <= 1:
        return n
    memo[n] = fibonacci(n-1) + fibonacci(n-2)
    return memo[n]

Security

Security-First Design

AgentOS is built with security at its core. Multiple layers of protection ensure your system stays safe while agents perform their tasks autonomously.

Command filtering blocks dangerous operations
Input validation prevents injection attacks
Optional Docker isolation
Configurable resource limits

Command Filtering

Blocks rm, sudo, dd, mkfs...

Input Validation

Sanitizes ; && || | $() ...

Docker Isolation

Sandboxed execution environment

Resource Limits

Timeouts & step limits

Management

Process Management

Monitor and control your agents with a powerful process management system. Track status, view logs, and manage agent lifecycle with simple commands.

SQLite-backed agent registry
Real-time status tracking
Comprehensive logging
Graceful shutdown support

$ agentos ps

assistant ● running create API

researcher ● completed analyze data

debugger ● failed fix bugs

Configuration

Simple YAML Manifests

Define your agents with clean, readable YAML configuration files. No complex setup required - just describe what you want and run.

Human-readable configuration
Version control friendly
Shareable agent definitions
Custom security rules per agent

agent.yaml

name: research_assistant
model_provider: claude
model_version: claude-3-5-sonnet
isolated: true

DESTRUCTIVE_COMMANDS:
  - rm
  - sudo
  - rmdir

And Much More

Additional features that make AgentOS production-ready

Retry Logic

Automatic retries for failed LLM calls with exponential backoff

Chat History

Persistent conversation history with context preservation

Rich CLI

Beautiful terminal output with colors and formatting

Extensible

Easy to add custom providers and tools

SQLite Backend

Lightweight, reliable data storage

Open Source

MIT licensed, free to use and modify

Features