5 Herald OS Features Shown in the 2026 Demo

Herald OS, as described in the video, is a free, open-source dashboard for managing AI agents, with the agent itself serving as the interface. The demo covers setup, daily overview, spaces, missions, memory, files, automations, settings, voice, and known rough edges.

What Is Herald OS According to the Video?

Herald OS is described in the video as a free, open-source, agent-native operating system that uses an AI agent as its interface, published on GitHub. The presenter, who runs the channel Julian Goldie SEO, credits a developer he names Luke and says the project was released shortly before the video appeared on 2026-10-08.

One caveat matters before anything else: this article could not verify a canonical repository for the project named in the video, so no repository link is provided. Everything below reports the presenter's first-hand demo. Treat installation steps and feature names as claims from the video, not from documentation.

The presenter's stated motivation is usability. He recalls that when the agent first launched around March or April, users operated it inside a terminal, which he calls messy. He positions Herald OS as a nicer-looking alternative to the agent's existing desktop app, while noting the desktop app supports useful plugins.

How Does Setup Work?

Setup, per the presenter, took about two minutes. He says he opened Claude, Anthropic AI assistant, and asked it to perform the installation, which completed the agent OS for him. That is a speaker's first-hand account of one setup on one machine, not a documented or independently timed figure.

He does not walk through manual installation in the video, so readers who prefer verifiable instructions should wait for the project's own documentation. His implied workflow, using a coding assistant to install an open-source project, is common but depends on the project's actual README, which this article cannot confirm.

What Does the Dashboard Show?

The dashboard greets the user by name, lists the day's schedule, and highlights items that need attention. The presenter says the schedule view draws on tasks scheduled inside Herald OS itself, and that a working view surfaces what to pick up next.

A "pick up where you left off" feature reads the names of recent documents, projects, chats, and calendar events, then suggests where to continue when you return after a break. The presenter says you can switch this suggestion feature on or off.

The dashboard also surfaces recent desktop screenshots and recent work so they appear as context. He demonstrates a chat area with preview and source panels, where you can attach files the agent can read before starting a conversation.

Why Does It Have Three Separate Spaces?

Herald OS ships with three spaces: ideas, work, and personal. The presenter switches between them in the UI and says you can add a new space from the general settings, which he suggests as a clean way to separate clients or companies.

He frames this as an answer to a recurring question in his community: people running multiple businesses or serving multiple clients want isolated contexts for their agent workspaces. A default space can be set in settings, and each space keeps its own section in the sidebar.

This is a workspace-separation feature at the UI level. The video does not demonstrate permission boundaries or data isolation between spaces, so it should not be read as a security claim.

What Do Missions, Memory, and Files Do?

Missions

Missions are multi-step tasks you hand to the agent. The presenter types an example mission, "build out a new website," and says the agent plans and executes it step by step. The missions panel lists active missions with their plans.

Memory

The memory panel lists everything the agent has stored about you. The presenter points out his own Telegram details appeared there, and warns viewers to check what the panel exposes before screen-sharing. That is a practical privacy note worth repeating: review stored memory before demos.

Files

The files section manages everything the agent can access, organized into recent, favorites, projects, downloads, shared, and deleted. It is a file browser scoped to the agent's permissions rather than a full system file manager.

How Do Automations and Multi-System Sync Behave?

The automations panel manages scheduled jobs with simple on and off switches. The presenter shows examples including a daily 6:00 a.m. to-do recap, daily Japanese practice, and morning inbox summaries. He says the same automations appear across three of his systems and appear to sync.

The three systems he runs are Herald OS, the agent's desktop app, and a separate agentic operating system that supports agents beyond the one in this demo. He reports the sync "seems to" work well across all three, which is his observation across his own machines, not a documented sync guarantee.

One feature failed for him: the Connections panel, meant to bring external apps and tools into the agent, did not sync. He attributes this to the project being young and open source, and says he would ask a coding assistant to fix or extend it.

What Can You Configure in Settings?

Settings cover models, runtimes, agents, appearance, and voice. The presenter shows a preferred model setting, with what he calls GPT-6 plugged in through a CLI, and a runtime selector that includes Docker, the container platform, among other ways to run the agent.

Other settings include the number of background agents, whether the system remembers preferences, an auto-approve toggle with an ask-before-sending option, and an activity history view. He calls activity history especially useful for spotting background jobs you forgot about.

General settings control full-screen launch, default space, default projects folder, do-not-disturb mode with its notification bell, sleep and screensaver behavior, and dictation. A plugins section lets you paste a plugin reference from GitHub and add it to modify the agent. A software section manages connected tools including Claude Code, Anthropic terminal coding agent, Codex, and LM Studio, the local model runner.

The voice section lets you switch the speech engine, choose between a local voice agent and a live one, and review available voice commands. Usage settings show token consumption, current plan, and token balance. Privacy, notifications, network, storage, and accessibility panels round out the list.

Should You Choose It Over the Desktop App?

The video's verdict is aesthetic and practical rather than measured: Herald OS looks better, feels cleaner, and suits people who dislike the desktop app's interface, in the presenter's opinion. He also flags the Connections panel failure and a minor UI clipping issue at the top of a panel as rough edges.

Because the project could not be verified against a canonical repository at the time of writing, treat adoption as at your own risk until documentation exists. If you try it, review the memory panel and activity history early, keep auto-approve off until you trust a workflow, and confirm which model endpoint your token usage depends on.

Frequently Asked Questions

• Is Herald OS free? The presenter says it is free and open source and available on GitHub. This article could not independently verify the repository, so confirm licensing yourself before relying on it.

• How long does Herald OS take to set up? The presenter reports about two minutes using Claude to run the installation. That is one person's account on one machine, not a benchmark.

• Does Herald OS replace the desktop app? The video frames it as an alternative interface, not a replacement. The presenter still runs both, plus a third agentic operating system, on the same automations.

• What did not work in the demo? The Connections panel failed to sync for the presenter, and one panel clipped slightly at the top. He expects fixes since the project is open source.

• Can I use voice with Herald OS? Yes, per the demo: a dictation toggle, a selectable voice engine, a choice between a local voice agent and a live one, and a list of voice commands in settings.

Source video