OpenHands v1.7.0 リリースOpenHands Releases v1.7.0

OpenHands, the open-source platform for AI software development agents, has published version 1.7.0, a release that focuses on interface flexibility and on how agents handle memory and credentials. For teams that rely on autonomous or semi-autonomous coding agents, small adjustments to model selection and state persistence can meaningfully affect day-to-day workflow, which is why this update is worth a closer look.

The most visible change concerns the model picker. In v1.7.0, the LLM selector is now always shown rather than being conditionally displayed, while agent-profile switching has been moved to the tools menu. Together these changes appear intended to make choosing a language model a first-class, always-available action, while relocating the less frequently used profile controls to a more suitable place. The practical effect is that users can more readily swap the underlying model for a given task without hunting through settings.

A second notable addition is a persistent agent memory toggle. Agent memory refers to an AI agent's ability to retain information—such as prior context, project details, or learned preferences—beyond a single interaction. By exposing this behavior as a toggle, OpenHands lets users decide whether an agent should carry state forward or start fresh. Persistent memory can improve continuity on long-running tasks, but providing an explicit on/off control is a sensible design choice, since retained context is not always desirable and can raise questions about relevance, privacy, and token cost.

The release also adds the ability to overwrite a secret value directly from the edit form. Secrets—API keys, tokens, and similar credentials—are commonly required so that agents can call external services or models. Updating an existing secret previously appears to have involved a more roundabout process; allowing an in-place overwrite from the edit form streamlines credential management. This is a modest but useful quality-of-life improvement for anyone maintaining several integrations.

Some background helps frame why these particular changes matter. OpenHands began as OpenDevin and is developed as a community-driven, open-source project, positioning itself as an alternative to proprietary AI coding assistants. It provides an environment in which agents can write and execute code, run shell commands, and interact with a browser to complete software tasks. Because the platform is broadly model-agnostic and can be pointed at a range of LLM providers, a prominent, always-visible model selector is more than cosmetic—it lowers the friction of matching a model to a task, whether that means favoring a cheaper model for routine work or a more capable one for complex reasoning.

The update also lands amid sustained activity across the AI coding agent space, where tools such as GitHub Copilot, Cursor, and Cognition's Devin compete on autonomy, reliability, and integration depth. Agent memory in particular has become a recurring theme across the sector, as developers look for ways to give agents durable context without overwhelming them with stale or irrelevant information. Approaches vary from simple session persistence to more elaborate retrieval systems, and exposing memory as a user-controlled setting reflects a pragmatic middle ground rather than a fully automated stance. Credential handling is a related concern, since agents that act on a user's behalf increasingly need managed access to external systems, making secret management an ongoing area of refinement.

Taken together, v1.7.0 reads as an incremental, workflow-oriented release rather than a sweeping redesign. The highlighted features—an always-visible LLM selector, relocated agent-profile switching, a persistent memory toggle, and editable secret values—are attributed in the changelog to contributor hieptl, reflecting the project's open-source, pull-request-driven development model. That cadence of frequent, contributor-led improvements is characteristic of active open-source tooling, where usability refinements often ship alongside larger capability work.

Users running OpenHands in production should review the official release notes for the complete list of changes, including any adjustments not summarized here, and test the new memory and secret behaviors against their own configurations before rolling them out. As with any release that touches credentials and persistent state, verifying expected behavior in a controlled setting is prudent.