Hermes Agent Upgrades: What Ships Free and What the 1,300 Clicks Hide
By Matt Cullerton, founder of Mavric and Apso
Hermes Agent is a free, MIT-licensed, self-hosted AI agent from Nous Research. A TikTok from @future.with.ai98 presents five "Hermes upgrades," but four of them (Mission Control, Oracle, a shared memory system and Search OS) do not exist in Hermes Agent. They are the creator's own build on top of it. The claim of "over 1,300 clicks a day" is a single peak day on a chart whose average is 271 clicks a day.
What is Hermes Agent?
"Hermes" is an overloaded name. Nous Research has shipped a Hermes model family for years: Hermes 3 405B on 24 August 2024, Hermes 4 on 26 August 2025 and Hermes-4.3-Seed-36B on 3 December 2025. Hermes Agent is a separate product, dated 25 February 2026 on the Nous Research releases page, which describes it as "An autonomous agent that lives on your server, remembers what it learns, and gets more capable the longer it runs."
The GitHub repository is MIT licensed and had 257.5k stars when checked. Its README lists a built-in learning loop, a terminal UI, messaging gateways for Telegram, Discord, Slack, WhatsApp, Signal and CLI, a cron scheduler, parallel subagents, and backends for local, Docker, SSH, Modal, Daytona and Vercel Sandbox. The official site installs it with curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash.
The video says the upgrades are so powerful it "feels illegal to use them for free." The base agent is free to use, with no license fee. You still pay for a model provider, and the five demonstrated add-ons do not ship with it.
Are the five upgrades part of Hermes Agent?
Four are not. Neither the Nous Research README nor the official site mentions Apollo, Mission Control, Oracle, news scanning, SEO article generation or memory shared with other AI tools. A community changelog at hermes-ai.net, which says it is unaffiliated with Nous Research, lists releases back to v2026.3.17 and contains none of those names either.
The add-ons come from Julian Goldie. His site agentos.guide describes "Read-along scripts that pair with the YouTube videos. Built for the AI Profit Boardroom community," and routes paid access to a Skool community called AI Profit Boardroom. The dashboard in the video greets "Good afternoon, Julian.", and its memory graph includes a node labelled "Julian Goldie."
Voice is the one partial exception. The hermes-ai.net changelog records release v2026.8.3 as adding "Streaming voice with barge-in, wake words, and hands-free control." The name "Apollo" is the creator's branding.
Does Hermes Apollo run on your ChatGPT subscription?
The video says Apollo "runs on the chat gpt subscription you're already paying for." The mechanism exists, but in a third-party repository. TheSmokeDev/hermes-talk is MIT licensed and describes "Realtime duplex voice for your own agent" that "rides your ChatGPT or X Premium subscription." Its default path uses codex login with model gpt-realtime-2.1, and an xAI path uses grok-voice-latest. It also lists GPT-Live, Gemini Live and ElevenLabs as alternatives. The word "Apollo" does not appear in it.
OpenAI's openai/codex README documents the sign-in: "Run codex and select Sign in with ChatGPT. We recommend signing into your ChatGPT account to use Codex as part of your Plus, Pro, Business, Edu, or Enterprise plan." It mentions no voice, audio or Realtime API. So a community tool reuses Codex credentials to reach a voice model, and OpenAI's own Codex docs describe that sign-in for coding work only.
A review site, AIToolsReview, claims Hermes Apollo "does NOT run on ChatGPT Plus/Pro subscriptions." That contradicts the hermes-talk README. The repository implements the behavior, so it carries more weight, but the conflict is unresolved.
Did the SEO agent really reach 1,300 clicks a day?
The video says the SEO agent helped "grow our website from zero to over 1300 clicks a day." The Google Search Console panel shown at the same moment reads "Clicks per day 1,313" and "Total clicks 46,612" for aiprofitboardroom.com, with a chart running from 2026-04-18 to 2026-10-06.
That window is 172 days. 46,612 clicks over 172 days averages 271 clicks a day. The 1,313 figure is a tooltip on one spike near the right edge of the chart, 4.85 times the average, with the line falling after it.
"Our website" is also six sites. The settled view lists aiprofitboardroom.com at 46,612 clicks, aisuccesslabjuliangoldie.com at 12,701, aimoneylabjuliangoldie.com at 3,162, juliangoldieaiautomation.com at 2,935, agentos.guide at 1,323 and bestaiagentcommunity.com at 702, under a banner of 67,386 clicks and 4.1M impressions. The "from zero" part cannot be checked, because the chart starts above zero and shows no earlier data.
The video never shows the SEO agent finishing a run. Its screen sits at "STEP 2 OF 4" with "0/5 WRITTEN" and "0 LIVE."
What Oracle's news cards get wrong
Oracle is pitched as a tool that finds the day's news, ranks it and turns it into a post with one click. Its top card, "Anthropic ships Claude Managed Agents multi-agent beta," is a real event. Anthropic announced Claude Managed Agents in public beta.
The angle Oracle attached is wrong. It reads "Build 1000-agent workflows in Claude Code." Anthropic's workflow runs docs set the 1,000-agent ceiling for Managed Agents workflow runs, alongside 64 threads at once, a 24-hour default run lifetime and 10 open runs per session. That limit belongs to Managed Agents. It has nothing to do with Claude Code.
The second card, "Anthropic starts massive training run for Claude 6," comes from a single account with 545 likes and calls itself a rumor in its own text. Anthropic's models overview lists Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5 as current, with no Claude 6. Oracle scored it one notch below the real launch and gave it the same "Publish to WP" button.
Bottom line
Hermes Agent itself is real, free and well documented. The five "upgrades" are mostly one creator's paid-community build, the ChatGPT voice trick depends on a third-party repo, and the 1,300 clicks a day is one peak day against a 271-a-day average across one of six sites.
Sources
- Nous Research Releases: Hermes Agent's release date and the separate Hermes model family.
- NousResearch/hermes-agent: license, star count and built-in features.
- Hermes Agent official site: shipped capabilities and install command.
- hermes-ai.net changelog: unofficial changelog with the v2026.8.3 voice release.
- TheSmokeDev/hermes-talk: the voice tool that reuses
codex login. - openai/codex: the "Sign in with ChatGPT" flow for Codex.
- Claude Managed Agents announcement: the public beta behind Oracle's top card.
- Workflow runs docs: the 1,000-agent ceiling and related limits.
- Multiagent orchestration docs: Managed Agents beta status.
- Claude models overview: the current Claude lineup.
- agentos.guide: Julian Goldie's guides and the paid community.
- AIToolsReview on Hermes Apollo: a review that conflicts with the hermes-talk README.
Published October 12, 2026. Based on a TikTok by @future.with.ai98. Full video notes (markdown)