With “Agent Canvas,” the open-source project OpenHands offers a self-hostable control center for AI coding agents. The MIT-licensed tool runs locally by default and turns individual coding agents into an always-on “engineering team.”
Start local, scale as needed
Agent Canvas is a “local-first” workspace where you can start one or multiple agent sessions, inspect conversations and manage files and terminal output. Useful patterns can be turned into automations – such as reports that publish to Slack, or automatically decomposing GitHub issues into tasks. By default everything runs on your own machine; when needed, agents can run in Docker containers, on VMs or within your own company infrastructure.
Open to third-party agents
Out of the box, Agent Canvas uses the open-source OpenHands agent, but it can also integrate third-party agents like Claude Code or Codex. That makes the tool a vendor-neutral interface for agentic programming.
Who it is for
For developers and teams who want to run AI coding agents productively and under their own control – without locking into a single cloud provider. The source code is freely available on GitHub.
