Julian Goldie pitches a paid 'agent OS' dashboard built on Nous Research's open-source Hermes Agent

The video names these parts of the dashboard:

• Model profiles: separate profiles, several running on free models (01:05).
• Bot screens: cloud desktops that keep running after the user shuts down (01:31).
• Named agents: Apollo (voice), Oracle (news), Muse and Astros (content ideas) (02:27).
• Other features: lead gen with "mixture of agents" (04:42), MCP and goal mode (05:19), and a Claude-plan plugin (06:10).

The video does not give a Hermes version number. To get the dashboard you join the "AI Profit Boardroom", which includes the zip file, a 30-day roadmap and four calls a week (02:57).

Most of the "while I sleep" features already exist in Hermes itself, according to its official repo (GitHub):

• Scheduling: a built-in cron scheduler.
• Parallel work: isolated subagents.
• Memory: persistent memory with FTS5 session search and summaries.
• Skills: agent-written skills that it curates itself.
• Model access: OpenRouter, OpenAI and custom endpoints.
• Where it runs: seven terminal backends, including Docker, SSH, Modal, Daytona and Vercel Sandbox. Cloud sandboxes can hibernate when idle.

The tracker lists Profiles and MCP server mode as arriving in v0.6.0 (Release Timeline). The Goldie dashboard looks like a UI and workflow layer on top of these features. Its memory "galaxy" is shown but not explained. Earlier coverage describes the memory as a plain-text Obsidian folder (OpenClaw Database). No code, configs or evaluations are public.

• Claim: the agents keep working while the user sleeps. This is plausible, because Hermes's cron scheduler and cloud backends are documented (GitHub). The dashboard's own behaviour is only shown, never measured.
• Claim: memory syncs across all agents. This rests only on the presenter's demo (06:20).
• Claim: the video was made by one click from idea to finished video. This is self-reported (03:40).
• Claim: 92,900 members in the free community. This is self-reported (07:48).
• No measured results: the video reports no figures for output quality, uptime or cost.

• Hermes Agent is described as the successor to OpenClaw, with a built-in learning loop (Agentic AI docs).
• GitHub star counts differ by date: about 180k in one profile (Agentic AI docs) and 253k on the repo now (GitHub).
• The video compares bot screens to "Grockbot" and ChatGPT's agent products, as heard in the auto-captions (01:47).
• This video continues a run of Goldie videos about a Hermes agent OS (AI Tools Review).

• The presenter warns that the list of free models "changes a lot" (01:26). Cost savings depend on how good free models are, which is not demonstrated (OpenClaw Database).
• The lead-gen tool can run outreach on its own. The presenter says he still reviews every message before it is sent (04:56).
• Sources disagree on the launch date (February vs. March 2026).
• The video uses urgency selling ("a window… won't stay open forever") (07:11).

• Overnight jobs, profiles, MCP and shared memory are available free in MIT-licensed Hermes. Teams can test them directly before paying for a packaged dashboard (GitHub).
• A real advantage of this setup: when model profiles are separate from memory, switching models is cheap.
• Agents that run unattended and send outreach need human approval gates and spending limits.
• A dashboard sold only as a zip file, with no source repo, raises questions about maintenance and security that the reviewed sources do not answer.