Deploy

Name: ConnectOnion
Author: ConnectOnion

Beta

Deploy your ConnectOnion agents to production. Docker, GCP, AWS - your choice.

Why deploy? Run agents 24/7, scale horizontally, integrate with existing infrastructure. Your agents become production services.

Two Deployment Options

Recommended

co deploy

Quick deployment to ConnectOnion Cloud. Managed hosting, no infrastructure to manage.

One command deployment
Automatic HTTPS
Re-deploy updates same URL

Self-Host

Full control with your own infrastructure. Use Docker, GCP, AWS, or any VPS.

Full control
Custom domains
Compliance requirements

60-Second Quick Start

Deploy your agent with the CLI - one command:

Terminal
bash
# Create an agent project
co create my-agent

# Navigate to the project
cd my-agent

# Deploy to ConnectOnion Cloud
co deploy
output
Deploying to ConnectOnion Cloud...
 
  Project: my-agent
  Source: /Users/me/my-agent
  Package: 12.3 KB (8 files)
  Env: /Users/me/my-agent/.env (3 keys)
 
Uploading package to https://oo.openonion.ai...
Deployment: a1b2c3d4
Building container on ConnectOnion Cloud...
  [1/100] status: deploying
  [2/100] status: running
 
Deployed!
Agent URL: https://my-agent-0x7a9f3b2c.agents.openonion.ai
Dashboard: https://o.openonion.ai/dashboard

co deploy Requirements

Initialized ConnectOnion project

.co/host.yaml (created by co create or co init)

Authenticated (co auth)

Configuration

.co/host.yaml
yaml
# .co/host.yaml
name: my-agent
entrypoint: agent.py
env: .env

Environment Variables

Environment variables from .env are securely passed to your agent:

.env
bash
# .env
OPENAI_API_KEY=sk-xxx
DATABASE_URL=postgres://...

Note: URL format is {project_name}-{your_address[:10]}.agents.openonion.ai. Re-deploying updates the same URL.

Template Deploys with Skills

Deploy a built-in template plus any combination of skills — no local project needed. The CLI creates a temporary project, bundles your skills into .co/skills/, deploys, and cleans up:

Terminal
bash
co deploy --template co-ai --name linkedin-agent \
  --skills ~/skills/linkedin-login ~/skills/linkedin-post-submit
output
Creating temporary co-ai project...
 
Deploying to ConnectOnion Cloud...
 
  Project: linkedin-agent
  Skills:
    ~/skills/linkedin-login -> .co/skills/
    ~/skills/linkedin-post-submit -> .co/skills/
 
Deployment: aea819bb
Building container on ConnectOnion Cloud...
 
Deployed!
Agent URL: https://linkedin-agent-0x7a9f3b2c.agents.openonion.ai

--name sets the project name and URL (default: {template}-agent) — different skill combinations run side by side instead of overwriting each other.

--skills takes one or more paths. A path that is itself a skill (contains SKILL.md) lands at .co/skills/{dirname}/; a directory of skills has its contents merged in.

The co-ai image ships Chrome + Xvfb + Node.js, so browser skills and their node scripts work out of the box.

Builds run in the background — if the CLI stops polling before a slow image finishes, the deploy continues server-side. Check progress with co status.

Choose a Dedicated Server Region

Use co server new for a separate machine, then deploy your agent onto it. Sydney remains the default. Melbourne offers another Australian location; Singapore, Iowa, and Oregon are also available when explicitly selected.

Terminal
bash
# Create one dedicated browser-agent server in Melbourne
co server new prod --machine e2-medium --region australia-southeast2

# Deploy your initialized agent project onto it
co deploy --to prod

Region IDs: australia-southeast1 (Sydney), australia-southeast2 (Melbourne), asia-southeast1 (Singapore), us-central1 (Iowa), and us-west1 (Oregon). The CLI reads available regions from the pricing API. Region selection does not migrate an existing server, and quota or capacity can still prevent creation.

Server credit is charged for 12 months up front: US$360 for e2-small or US$660 for e2-medium, with the same price in every listed region. Review the live quote before confirming. Failed provisioning reports whether the full refund succeeded; do not retry an uncertain purchase until its status is checked.

Self-Host Paths

VPS / bare metalPython + host() →Docker / containerAny Docker host →Google CloudCloud Run →AWSEC2 or ECS →

Self-Host with host()

Deploy to your own infrastructure using host():

agent.py
python
# agent.py
from connectonion import Agent, host

agent = Agent("my-agent", tools=[my_tool])

# Export ASGI app for uvicorn/gunicorn
app = host.app(agent)

if __name__ == "__main__":
    host(agent)

Run with uvicorn/gunicorn

code
bash
# Direct
python agent.py

# Uvicorn
uvicorn agent:app --workers 4

# Gunicorn
gunicorn agent:app -w 4 -k uvicorn.workers.UvicornWorker

For full API reference, see host() documentation.

Docker Deployment

The simplest way to deploy - works anywhere Docker runs:

1. Create Your Agent

agent.py
python
# agent.py
from connectonion import Agent, host

def search(query: str) -> str:
    """Search for information."""
    return f"Results for: {query}"

agent = Agent(
    name="my-agent",
    tools=[search],
    system_prompt="You are a helpful assistant."
)

# Host the agent
host(agent)

2. Dockerfile

Dockerfile
dockerfile
FROM python:3.11-slim

WORKDIR /app

# Install dependencies
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Copy agent code
COPY agent.py .

# Run the agent
CMD ["python", "agent.py"]

3. Build & Run

Terminal
bash
# Build the image
docker build -t my-agent .

# Run with API key
docker run -d \
  -e OPENAI_API_KEY=$OPENAI_API_KEY \
  --name my-agent \
  my-agent
output
Building image...
Successfully built a1b2c3d4e5f6
Successfully tagged my-agent:latest
 
Container started: my-agent
Agent serving at: 0x3d4017c3e843895a...

Deploy to Google Cloud Run

Serverless deployment with automatic scaling:

1. Build & Push to Container Registry

Terminal
bash
# Authenticate with GCP
gcloud auth login

# Set project
gcloud config set project YOUR_PROJECT_ID

# Build and push
gcloud builds submit --tag gcr.io/YOUR_PROJECT_ID/my-agent

2. Deploy to Cloud Run

Terminal
bash
gcloud run deploy my-agent \
  --image gcr.io/YOUR_PROJECT_ID/my-agent \
  --platform managed \
  --region us-central1 \
  --set-env-vars "OPENAI_API_KEY=sk-..." \
  --allow-unauthenticated
output
Deploying container to Cloud Run service [my-agent]...
Done.
 
Service URL: https://my-agent-abc123-uc.a.run.app
Agent address: 0x3d4017c3e843895a92b70aa74d1b7ebc...

Security Note

Use Secret Manager for API keys in production instead of environment variables.

Deploy to AWS

Multiple options depending on your needs:

EC2 / Lightsail

Simple VPS deployment. Best for always-on agents.

Full control over environment
Predictable pricing
Easy SSH access

ECS / Fargate

Container orchestration. Best for scaling.

Auto-scaling built in
Load balancing
Rolling deployments

EC2 Quick Deploy

Terminal
bash
# SSH to your EC2 instance
ssh -i your-key.pem ubuntu@your-instance-ip

# Install dependencies
sudo apt update && sudo apt install -y python3-pip docker.io

# Clone and run
git clone https://github.com/your/agent-repo.git
cd agent-repo
docker build -t my-agent .
docker run -d -e OPENAI_API_KEY=$OPENAI_API_KEY my-agent

Environment Variables

Configure your agent for different environments:

Variable	Description	Required
OPENAI_API_KEY	OpenAI API key for GPT models	Yes*
ANTHROPIC_API_KEY	Anthropic API key for Claude	Optional
GOOGLE_API_KEY	Google API key for Gemini	Optional
RELAY_URL	Custom relay server URL	Optional
LOG_LEVEL	Logging verbosity (DEBUG, INFO, WARN)	Optional

* Or use ConnectOnion managed keys with co auth - no API keys needed!

Best Practices

Security

Use secret managers for API keys
Never commit .co/ or .env
Use non-root container users
Enable HTTPS for all endpoints

Reliability

Add health checks to containers
Set up automatic restarts
Configure logging and monitoring
Use persistent volumes for keys

Ready to Deploy?

Your agents are production-ready. Ship them!

host() Reference Connect to Agents

Getting Started

Vibe Coding

CLI

Deploy

Two Deployment Options

co deploy

Self-Host

60-Second Quick Start

co deploy Requirements

Configuration

Environment Variables

Template Deploys with Skills

Choose a Dedicated Server Region

Self-Host with host()

Run with uvicorn/gunicorn

Docker Deployment

1. Create Your Agent

2. Dockerfile

3. Build & Run

Deploy to Google Cloud Run

1. Build & Push to Container Registry

2. Deploy to Cloud Run

Security Note

Deploy to AWS

EC2 / Lightsail

ECS / Fargate

EC2 Quick Deploy

Environment Variables

Best Practices

Security

Reliability

Ready to Deploy?

Star us on GitHub