OpenClaw Assistant runs an AI agent inside Home Assistant OS as an add-on

An add-on, not a plugin, and that difference decides the host

OpenClawHomeAssistant contains a Home Assistant add-on that runs OpenClaw inside Home Assistant OS. That single sentence sets the boundary of the project, and it is the first thing to check before reading any feature.

The add-on model is a Home Assistant OS concept. An add-on is a packaged application with its own container, configuration and lifecycle, installed from the add-on store, and it only exists on a Home Assistant installation running the operating system image. A Home Assistant installed as a container, or the Core installation, does not have the add-on store in the same way, so the installation path in the README does not apply to those setups.

The install itself is four steps through the user interface. In Home Assistant you go to Settings, then Add-ons, then the add-on store, then the repositories menu, and you add the repository URL. Then you install the add-on called OpenClaw Assistant.

The naming history is worth a note because it affects what you search for. The README records the upstream rename history as clawdbot, then moltbot, then openclaw as the final name, so older references, forum posts and screenshots may use the earlier names.

The repository is structured the way an add-on repository is, with a repository.yaml describing the add-on, an openclaw_assistant directory for the add-on itself, and the documentation files alongside.

What actually runs inside the add-on

The key features list is worth reading as a description of the running system rather than as a list of conveniences.

The AI Gateway is the OpenClaw server with chat, skills and automation capabilities, and it is the centre of the add-on. The Web Terminal is a browser-based terminal embedded in Home Assistant, so you get a shell on the machine without opening an SSH session. The Assist Pipeline feature lets you use OpenClaw as a conversation agent through the OpenAI-compatible API, which is the piece that makes the add-on useful to the rest of Home Assistant rather than a separate application that happens to be hosted there. Browser Automation comes with Chromium included for web scraping and automation skills, so the agent has a real browser rather than a simulated one.

Two operational features follow. Proxy Support is an optional outbound http_proxy setting for HTTP and HTTPS traffic, which matters in a home network behind a filtering resolver. Persistent Storage means skills, config and workspace survive add-on updates, and that is the right default for an agent whose configuration and learned skills take time to build.

The Bundled Tools line is a list of what you get in that terminal: git, vim, nano, bat, fd, ripgrep, curl, jq, python3, pnpm and Homebrew. That is the toolchain of a developer machine inside your home server, and it explains why the add-on is not a minimal container.

Three architectures, and what that says about the target hardware

The supported architectures table is short and gives a clear picture of the audience. amd64 is supported, aarch64 covers the Raspberry Pi 4 and 5, and armv7 covers the Raspberry Pi 3.

Supporting armv7 is the notable entry. That architecture is a 32-bit target, which rules out a lot of modern software, and the fact that it is there means the add-on is intended to run on the older and cheaper boards as well as on x86 servers. The practical implication is that a build exists for the hardware people actually have in a home, including the Pi 3, which is a machine that has already outlived several of its would-be replacements.

The bundled tool list is relevant here, because some of those tools are Go binaries and some are scripts, and supporting three architectures means all of them are built per platform. A bundled Chromium is the other part of that story, since it is the largest single dependency in the list and the one most likely to be the constraint on an armv7 device.

If you are running this on a Raspberry Pi, the architecture support is the good news and the browser is the question to ask about before you install.

The separate integration repository, and the split of responsibilities

The README announces at the top that the OpenClaw Home Assistant integration is available now, and gives a separate repository for it. That distinction is the thing to understand about the project as a whole.

This repository is the add-on. It runs the agent, the gateway, the terminal and the browser, and it exposes the OpenAI-compatible API. The separate integration repository is the piece that talks to Home Assistant proper, where the entities and the conversation pipeline live.

The split follows from what the add-on can and cannot do on its own. An add-on is a container on the Home Assistant host, and it can reach the Home Assistant API as a client with its own credentials, but it does not become part of the entity model. The Assist Pipeline feature is the bridge in the other direction, letting Home Assistant treat OpenClaw as a conversation agent over an OpenAI-compatible API, which is the standard way to plug an external model into the assist system.

So if your goal is an agent that can see and control your home, you will want the integration repository as well. If your goal is an agent with a browser, a terminal and its own persistent workspace, this repository is the whole of it.

Documentation, security, and what the add-on can reach

Two documentation files carry most of the substance, and the README links both directly.

DOCS.md is the full documentation, covering installation, configuration, use cases and troubleshooting. SECURITY.md is a separate document for the security risks and a disclaimer, and the README describes it as important risks to understand before using this add-on. Putting a security document in the front page of the README rather than in a wiki is a deliberate choice, and it tells you the author considers the risk surface part of the product rather than an appendix.

The risk surface is not small, and it follows from what the add-on bundles. There is a Chromium for web scraping, a shell with a developer toolchain, an outbound proxy option, and persistent storage of skills and configuration. An add-on with a browser and a terminal on a machine that controls your house is a large amount of capability in one container, and the Chromium in particular is the component that reaches out to the internet on your behalf.

The support channel is a Discord server, and the repository also carries AGENTS.md and CLAUDE.md, the instruction files for automated coding assistants, plus a .github directory and a .vscode directory. The MIT licence sits in the LICENSE file at the root.

The release cadence, and where the project is heading

The release history shows a project shipping often. The versions are v0.5.95 from 2026-10-16, v0.5.94 from 2026-10-08 and v0.5.93 from 2026-09-26, and the last push to the repository was on 2026-10-20 with the repository not archived.

A version number in the mid nineties of a zero major version, moving every week or two, is the signature of a project that releases continuously rather than on a schedule. The versions are 0.x, so nothing is promised about interface stability, and with releases this frequent the changelog and the release notes are where the actual change history lives rather than the README.

The primary language of the repository is Shell, which is consistent with a project that is mostly packaging: the Dockerfiles, the run scripts and the configuration that make OpenClaw work inside a Home Assistant add-on. The real complexity of the project sits in the upstream software it ships, and this repository's job is to make that software installable and configurable in one place.

The rename history is the other signal. clawdbot became moltbot became openclaw, and an add-on that follows an upstream project through renames is going to keep renaming its add-on identifiers and configuration keys as upstream settles.

Where OpenClaw Assistant is the wrong choice

The first limit is the host. The add-on runs inside Home Assistant OS, and the installation path in the README goes through the add-on store, so a Home Assistant container or a Core install is a different problem. The add-on is not a Docker image you compose yourself, even though the internals are containerised.

The second is the weight. An add-on that bundles Chromium, a shell with git and vim and python3 and pnpm, and Homebrew, is not a small thing to run on a Raspberry Pi 3, which is the weakest architecture on the support list. It is also a large attack surface, and the project's own SECURITY.md exists to make that explicit. If your Home Assistant install is a small box and you want it to stay quiet, this is the wrong add-on.

The third is what the agent can do. OpenClaw has chat, skills and automation capabilities and a browser for scraping, which is a broad mandate on a machine that sits inside your home. The security document is the place to check what that means concretely before you install it, rather than after.

The fourth is the Home Assistant integration. This repository does not give you entities, automations or a conversation pipeline in Home Assistant. That is the separate integration repository, and if your goal is the assistant integration rather than the agent runtime, you are looking at the wrong repository.

The fifth is the version churn. Releases every few days at 0.x means you will be updating the add-on often, and while the persistent storage protects your skills and config across updates, anything you configure outside that stored area is yours to manage.

Editorial conclusion

OpenClaw Assistant fits a Home Assistant OS user who wants an AI agent with its own browser and terminal living inside the same system as their home automation, and who is comfortable with an add-on that pulls in a Chromium and a full toolchain. It is the wrong tool if you run Home Assistant in a container or on Core rather than HAOS, since the add-on model depends on the add-on store, and it is the wrong tool if you want a tight integration with Home Assistant entities, because that is the separate integration repository rather than this one. Before you install it, check three things: that your system is Home Assistant OS, since the add-on runs inside HAOS, that you read the SECURITY.md file the README points to, because the add-on ships a browser and an outbound proxy option, and that you understand the persistent storage behaviour, since skills, config and workspace survive add-on updates and that is a deliberate trade. The code is MIT licensed, the current release is v0.5.95 from 2026-10-16, and the last push was on 2026-10-20.

Frequently asked questions

Official sources

1. Issues
2. License: MIT
3. README
4. Releases
5. techartdev/OpenClawHomeAssistant on GitHub