Intel / repo

OpenHands v1.28.0

Source: github.com
Author: openhands-release-bot[bot]
Date: Published Oct 19, 2026

Context

OpenHands v1.28.0 is mostly a polish release (GitHub). More than half of the listed bug fixes make themed text and error messages readable, fit dialogs on phone screens, or tidy the automations editor. The most notable change to conversations is that the can now ask for secrets. A secret here means a credential, such as an API key or access , that the agent needs to work with another service. One change lets the agent request a secret through a modal dialog inside the conversation. Another asks for requested secrets inline in the chat.

Two fixes target conversations that run in the cloud. A message sent before the cloud runtime is reachable is now held instead of failing. Resume now wakes a paused cloud instead of calling a runtime that isn't running. The release notes list only pull-request titles, so they don't say how a requested secret is stored or whether the model can see its value.

Key takeaways · AI-distilled

OpenHands 1.28.0 holds a chat message until the cloud runtime is reachable instead of failing it, and Resume now wakes a paused cloud sandbox rather than calling its runtime directly.
The release mounts the MARS web bridge by default, dropping the MARS_WEB=1 opt-in, and bumps the bundled agent-server to 1.55.1 and the automation component to 1.21.0.
Several fixes change failure handling: settings saves no longer retry a 4xx answer, failed profile and automation saves now show the server reason, and launch fails if the automation backend exits during startup.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.

Read the source github.com