What it does
OpenHands is an MIT-licensed project whose current README presents Agent Canvas, a front end for starting conversations with coding agents and setting up recurring tasks. By default everything runs on your own computer, and the same interface can switch between agent servers running locally, in Docker containers, on virtual machines or inside company infrastructure. Agent Canvas runs the OpenHands agent out of the box. It can also drive third-party agents such as Claude Code, Codex, Gemini CLI, Pi and OpenCode, or any agent that speaks the Agent-Client Protocol, and it works with whichever LLM you bring. Automations can run on a schedule or in response to webhook events, and they connect to services such as Slack, GitHub, Linear and Notion. The page gives examples like reports that publish to Slack and GitHub issues that are broken down into tasks, and it suggests sharing one agent server with a team for code review and dependency updates. You can install it with npm, run it in a Docker container, or build it from source. The page warns that running without a sandbox gives the agent full access to your filesystem. It names OpenHands Cloud and OpenHands Enterprise as commercial options for running agents on hosted infrastructure.
Capabilities
- Runs the OpenHands agent plus third-party agents such as Claude Code and Codex
- Switch between local, remote and cloud agent backends from one interface
- Scheduled and webhook-triggered automations
- Integrations with Slack, GitHub, Linear and Notion
- Optional Docker sandbox, including one container per conversation
- Works with any LLM you supply
When to use it
- You want to run several coding agents from one self-hosted interface
- You want agents to handle recurring work such as code review or dependency updates
- You need agents to run on a server or in containers rather than only on a laptop
