Operator OS is an ultra-lightweight, high-performance personal AI Agent framework written in Go. Designed to run on hardware as inexpensive as $10 with a memory footprint of less than 10MB, Operator OS brings true continuous intelligence to the edge.
Built from the ground up to support autonomous agents, persistent memory, and multi-channel messaging, it bridges the gap between complex reasoning models and constrained runtime environments.
- Ultra-Lightweight Engine: Consumes <10MB of RAM—99% smaller than typical Node.js or Python-based agent frameworks.
- Lightning Fast Boot: Cold starts in under 1 second, even on single-core 0.6GHz processors.
- True Portability: Deploys as a single, self-contained binary across RISC-V, ARM, and x86 architectures.
- Continuous Memory: Structural, persistent long-term memory that carries context seamlessly across sessions and reboots.
- Multi-Channel Integration: Natively supports Slack, Discord, Telegram, and WhatsApp out of the box.
- Universal Model Support: Drop-in support for OpenAI (5.x), Anthropic (Claude 4.x), Google Gemini (3.x Pro), and local Ollama.
Download the appropriate binary for your system from our Releases page.
Requires Go 1.21+.
git clone https://github.com/avrilonline/Operator-OS.git
cd Operator-OS
make deps
make build1. Initialize your workspace
operator onboard2. Configure your API keys
Edit ~/.operator/config.json to link your preferred LLM and messaging channels:
{
"model_list": [
{
"model_name": "gpt-5.2",
"model": "openai/gpt-5.2",
"api_key": "sk-your-openai-key"
},
{
"model_name": "claude-4-5-sonnet",
"model": "anthropic/claude-4-5-sonnet-20260220",
"api_key": "sk-ant-your-key"
},
{
"model_name": "gemini-3.1-pro",
"model": "gemini/gemini-3.1-pro",
"api_key": "AIza-your-google-key"
}
],
"agents": {
"defaults": {
"model": "claude-4-5-sonnet"
}
},
"channels": {
"slack": {
"enabled": true,
"bot_token": "xoxb-YOUR_SLACK_BOT_TOKEN",
"app_token": "xapp-YOUR_SLACK_APP_TOKEN"
}
}
}3. Start the Gateway Run the Gateway daemon to bring your agent online and connect it to your configured channels.
operator gatewayYou can now message your agent directly via Slack (or your chosen channel) or interact with it via the CLI:
operator agent -m "What is the status of the system?"Operator supports a zero-code model configuration system. Simply prefix the model name with the provider protocol in your model_list.
| Provider | Protocol Prefix | Example |
|---|---|---|
| Anthropic | anthropic/ |
anthropic/claude-4-5-sonnet-20260220 |
| OpenAI | openai/ |
openai/gpt-5.2 |
| Google Gemini | gemini/ |
gemini/gemini-3.1-pro |
| Groq | groq/ |
groq/llama3-8b-8192 |
| Ollama (Local) | ollama/ |
ollama/llama3 |
By default, Operator OS runs its agents in a secure sandbox.
- Workspace Confinement: Agents are restricted to reading/writing files within the configured workspace directory (default:
~/.operator/workspace). - Command Filtering: The
exectool blocks dangerous system commands (rm -rf, disk formatting, system shutdowns) even if workspace restrictions are bypassed.
To grant the agent full access to your host system (e.g., for system administration workflows), you must explicitly disable the sandbox in your configuration:
{
"agents": {
"defaults": {
"restrict_to_workspace": false
}
}
}Note: Only disable sandboxing in trusted, single-user environments.
To run the Gateway completely containerized without installing Go locally:
git clone https://github.com/avrilonline/Operator-OS.git
cd Operator-OS
# Generate the default configuration structure
docker compose -f docker/docker-compose.yml --profile gateway up
# Edit the generated config file with your keys
vim docker/data/config.json
# Start the Gateway in the background
docker compose -f docker/docker-compose.yml --profile gateway up -dWe welcome pull requests and issues! As Operator OS expands, we are looking for community maintainers to help build new skills, channel integrations, and deployment guides for edge hardware.
See the Roadmap for our current trajectory.
This project is licensed under the MIT License.



