Deploying powerful coding agents is easy, but managing them across different environments is a logistical nightmare. As you scale your use of LLMs, keeping track of which agent runs on which backend—and ensuring they can interact with your tools safely—becomes an impossible task.
OpenHands offers a definitive solution. It acts as a self-hosted developer control center designed specifically for orchestrating coding agents and automations across local, remote, and cloud backends.
A Unified Control Center for Your Engineering Team
The core strength of OpenHands lies in its ability to transform individual agents into an always-on engineering team. By utilizing the OpenHands repository, developers can run a variety of agents including Claude Code, Codex, and Gemini via any Agent-Client Protocol (ACP)-compatible interface.
The platform allows you to switch between different backends without losing focus. Whether you are running an agent on your local machine for quick tasks or utilizing a dedicated server for long-running processes, the Agent Canvas provides a seamless frontend experience. You can even configure complex prebuilt automations to handle repetitive tasks like generating Slack reports or decomposing GitHub issues.
- Self-host your way: Run agents in Docker, on VMs, or anywhere you can host an agent server.
- Deep Integration: Connect workflows with Slack, GitHub, Linear, and Notion.
- Bring Your Own Model: Full compatibility with any LLM of your choice.
- Flexible Backends: Easily move between local, remote, and cloud environments.
Deployment Strategies: Sandbox vs. Local
When setting up OpenHands, you face a critical decision regarding security and isolation. Depending on your needs, you can run the agent server directly on your host or within a controlled environment.
| Deployment Method | Security Level | Best Use Case |
|---|---|---|
| Without Sandbox | Low (Full Filesystem Access) | Quick local testing and prototyping. |
| Docker Sandbox | High (Isolated Environment) | Production workloads and untrusted code execution. |
Option 1: Rapid Local Setup
If you are working in a trusted environment, you can install the package globally using npm. Note that this method gives the agent full access to your filesystem.
npm install -g @openhands/agent-launcher
agent-canvasOption 2: Secure Docker Implementation
For more robust workflows, running within a Docker sandbox is recommended. This allows you to define exactly which project folders the agent can access. For those looking for advanced integration, exploring Giving Agents Real Power: A Guide to MCP Servers can help you further expand the toolset available to your agents.
To set up a Docker sandbox on macOS or Linux, use the following configuration:
export PROJECTS_PATH="$HOME/projects"
mkdir -p "$PROJECTS_PATH" "$HOME/.openhands"
docker run -it --rm \
-p 8000:8000 \
-v "$HOME/.openhands:/home/openhands/.openhands" \
-v "${PROJECTS_PATH}:/projects" \
ghcr.io/openhands/agent-canvas:1.0.0-rc.11
Scaling Your Infrastructure
The architecture of OpenHands is designed for scalability. The Agent Server acts as a REST API that can run multiple agents on a single machine, while the Automation Server enables scheduled or event-driven workflows. This setup is particularly useful when configuring backend setups on virtual machines to ensure your agents stay active even after you close your laptop.
For developers concerned about security, it is vital to remember that running untrusted code requires strict isolation, much like the principles found in E2B sandboxes.
Take Control of Your AI Workflow
Stop managing fragmented scripts and start managing a unified engineering team. Whether you deploy locally or in the cloud, OpenHands provides the control center your modern AI workflow demands.
Ready to automate your development lifecycle? Deploy OpenHands on your server today!