What OpenHands is
OpenHands gives a coding agent a place to work on a repository and a person a place to inspect that work. Its current Agent Canvas interface can connect to different execution backends, so the visible application and the machine running commands are separate choices.
Agent Canvas is the interface for following work and connecting an agent to a place where commands run. A Docker backend can give a task a contained working environment with selected project files, while direct local execution runs with the host’s access. This is developer software that needs configuration and review, rather than a ready-made assistant with every account connected.
Read the official documentationHow it fits into a workflow
Start with a repository task that has a concrete acceptance check, such as a reproducible bug. Give the agent only the relevant project access, inspect its changes, and review the actual command output. A proposed pull request is the beginning of review, rather than proof that the problem has been resolved.
Examples described as illustrative or proposed trials are editorial suggestions. Documented real builds, when available, are linked separately below.
Requirements, cost and limitations
The direct setup requires Node.js 24 or later and uv; the Docker route requires Docker and an explicit project-directory mount. Configure a model provider and the tools needed by the repository. The official VM guide suggests 2 virtual CPUs and 4 GB RAM for one user, which sizes the agent host rather than a local language model.
As checked on 30 October 2026, the project code is MIT-licensed, while model calls, hosted services and a cloud VM have separate costs. No standard cost per task is established. Direct mode gives broad filesystem and command access, and a reachable agent server can expose those capabilities to whoever can access it. Keep the execution environment scoped to the intended work.
Maintenance and version context
2026-10-27: v1.34.0 release documents per-conversation Docker runtime work and fixes.
When to choose it and what to compare
This is a useful option when the work includes repository investigation and repeatable engineering actions. Consider the execution backend, model route and review process together. A beginner can evaluate one small task before enabling scheduled workflows or broader access to projects and external services.
Technical details and documented capabilities
These details come from the linked sources, checked 2026-10-30. Open the current documentation before planning a deployment or migration.
- Type
- Community-built agents
- Runs on / through
- macOS, Windows, Linux
- Execution choices
- local, self-hosted, cloud
- License or access terms
- MIT
- Agent Canvas separates the interface from the agent execution backend.
- The open-source OpenHands agent runs alongside support for other ACP-compatible agents.
See an implemented example.
REAL IMPLEMENTATION / Graham Neubig / OpenHandsProposing GitHub issue fixes with OpenHandsFocused source review : Current repository, license, setup, permissions and published implementation evidence; not independently executed. Documentation review, not hands-on product testing.
Sources and evidence
Product facts are based on the following published sources, checked 2026-10-30. Workflow suggestions and comparisons are editorial interpretation; PrisAIm has not independently benchmarked this product.
Current product and setup MIT license VM sizing and security Release history Historical resolver implementation