OpenHands v1.10.0 リリースOpenHands Releases v1.10.0

OpenHands, the open-source platform for autonomous software development agents, has published version 1.10.0, a release that bundles several user-facing features spanning model configuration, observability, and automation discovery. The update matters because it touches components that developers interact with regularly, from the default model powering the Canvas workspace to the way teams audit agent activity and surface prebuilt automations.

The headline change is that Canvas now defaults to GLM 5.2. Canvas is OpenHands' interactive building surface, and switching its default backend signals that the maintainers consider GLM 5.2 a strong general-purpose choice for that workflow. GLM refers to the large language model family developed by Zhipu AI, also known as Z.ai, which has shipped successive versions positioned as competitive open-weight options for coding and agentic tasks. Setting GLM 5.2 as the default does not necessarily prevent users from selecting a different model; in OpenHands, default assignments are typically configurable, so the change appears to be about picking a sensible starting point rather than locking users into a single provider.

The release also adds activity log export, contributed alongside the other features in this cycle. An export capability of this kind is useful for auditing what an agent did, sharing a record with teammates, or feeding logs into external analysis. For teams running autonomous or semi-autonomous coding agents, being able to extract a durable trail of actions is a practical step toward reproducibility and accountability, which are recurring concerns as agents take on more of the work in a repository.

Another new addition is a featured automations landing dashboard. This appears to be a curated entry point that highlights ready-made automations, lowering the barrier for users who want to adopt existing workflows rather than build them from scratch. A discovery surface like this is a common pattern in tools that ship reusable components, since it helps users understand what is possible and reduces the time spent searching. The feature complements OpenHands' broader direction of packaging repeatable agent behavior into shareable units.

The changelog additionally references work on the skills page, including filtering, though the source excerpt is truncated at that point. Skills in agent platforms generally represent modular capabilities the agent can invoke, and adding filtering suggests the catalog has grown large enough that navigation and search have become worthwhile investments. As with the automations dashboard, the emphasis here is likely on making a growing library easier to browse.

For readers unfamiliar with the project, OpenHands is a framework for AI agents that can write code, run commands, browse, and interact with a development environment to complete software tasks. It has become one of the more visible efforts in the open-source agent space, sitting alongside a broader industry push toward autonomous and assistive coding tools. Canvas, skills, and automations together reflect a design philosophy that treats agent capabilities as composable pieces users can configure, reuse, and share, rather than a single monolithic assistant.

The choice of GLM 5.2 as a default is worth situating in a wider trend. Model selection has become a meaningful product decision for agent frameworks, because the underlying model shapes cost, latency, and task quality. Open-weight families from providers such as Zhipu AI have gained attention as alternatives to proprietary offerings, and defaulting to one of them keeps the platform aligned with an open ecosystem. It also means users benefit from improvements in the model line without changing their own configuration, provided they stay on the default.

As a minor version bump from the 1.9 line, v1.10.0 looks incremental rather than a major architectural shift, focusing on refinements to existing surfaces. Users typically upgrade by pulling the latest release through their existing installation method, and full attributions for each change are listed in the project's release notes on its repository. Anyone relying on the previous Canvas default may want to confirm the GLM 5.2 switch behaves as expected for their use case, and to check the complete changelog for the remaining items that the summarized excerpt does not fully capture, including the skills page filtering work and any smaller fixes shipped in the same cycle.