OpenHands: AI-Driven Development with Agent Canvas

Introduction

OpenHands Agent Canvas is a powerful, self-hosted developer control center for coding agents and automations. It transforms your coding agents into an always-on engineering team, enabling you to start conversations and automate tasks like generating reports or decomposing GitHub issues. The platform supports running OpenHands, Claude Code, Codex, Gemini, or any ACP-compatible agent across various backends, from local machines to cloud infrastructure.

Installation

OpenHands Agent Canvas can be installed and run on various machines, including laptops, dedicated computers, or cloud servers. Running it on a server in the cloud is recommended for continuous operation and easier integration with third-party services.

Option 1: Without a Sandbox

This method runs the agent-server directly on your machine, granting it full filesystem access.
Prerequisites: Node.js 22.12.x or later, uv

npm install -g @openhands/agent-canvas
agent-canvas

The agent-canvas command starts the full local stack. You can also run frontend or backend separately:

agent-canvas --frontend-only  # static frontend + ingress only
agent-canvas --backend-only   # agent server + automation backend + ingress only

Option 2: With a Docker Sandbox

This option provides a sandboxed environment using Docker.
Prerequisites:

• Docker: Docker Desktop on macOS/Windows, or Docker Engine/Docker Desktop on Linux.
• A host directory for PROJECTS_PATH containing project folders the agent needs to access. Create it before starting the container.

macOS / Linux:

export PROJECTS_PATH="$HOME/projects"  # directory containing your project folders
mkdir -p "$HOME/projects" "$HOME/.openhands"

docker run -it --rm \
  -p 8000:8000 \
  -v "$HOME/.openhands:/home/openhands/.openhands" \
  -v "${PROJECTS_PATH}:/projects" \
  ghcr.io/openhands/agent-canvas:1.12.0

The agent will be able to access any project under PROJECTS_PATH.
Access the UI at http://localhost:8000 for npm/source launchers, or http://localhost:8000/canvas for the Docker image. Additional backends can be added from the UI.

Examples of Use

OpenHands Agent Canvas excels at automating development workflows and integrating with existing tools. You can create automations and workflows that interact with services like Slack, GitHub, and Linear. For instance, agents can be configured to generate reports and publish them to Slack, or automatically decompose GitHub issues into smaller, manageable tasks. This enables a seamless integration of AI agents into your daily development operations, enhancing productivity and consistency.

Why Use OpenHands Agent Canvas?

OpenHands Agent Canvas offers several compelling reasons for developers and teams:

• Self-Hosted Flexibility: Run agents locally, in Docker, on VMs, or within your company infrastructure, maintaining full control over your data and environment.
• Backend Versatility: Seamlessly switch between local, remote, and cloud agents without losing context, adapting to your specific needs.
• Powerful Automations: Create sophisticated automations and workflows that integrate with popular services like Slack, GitHub, and Notion, running on schedules or in response to webhook events.
• LLM and Agent Agnostic: Use with any Large Language Model (LLM) and any agent compatible with the Agent-Client Protocol (ACP), including OpenHands, Claude Code, and Codex.
• Developer Control: Provides a centralized control center for managing and orchestrating your AI coding agents, enhancing efficiency and consistency in development tasks.

Links

For more information, documentation, and community engagement, please refer to the official links:

• GitHub Repository
• Documentation
• Join the Slack Community