Hermes Agent for content creation earns a 7.2 out of 10 from me: excellent at audio, voice capture and privacy, held back by setup effort and the lack of any built-in hold before something gets published.
The scores below are my editorial reading of the official v0.23.0 docs and plugin catalog, criterion by criterion.
Each rating carries a one-line reason you can check against the linked source.
Planning content that has to rank? Book a free SEO strategy session with my team before you pick a tool.
The Verdict Up Front
- Best at: turning one source, like a video or a set of notes, into many drafts in a voice that sounds like you.
- Strongest add-ons: the official ElevenLabs Music, ElevenLabs Sound Effects and Meshy plugins.
- Weakest point: publishing safety is something you configure, not something you get.
- Right for: owners and small teams willing to spend a few evenings on setup.
Scorecard: Nine Criteria Out Of 10
| Criterion | Score | Why, from the docs |
|---|---|---|
| Writing in your voice | 8 | /learn builds a skill from your files or URLs; humanizer strips AI-isms |
| Research and sourcing | 8 | web_search up to 100 results, web_extract reads pages and PDFs |
| Repurposing | 8 | youtube-content turns transcripts into summaries, threads and blogs |
| Images and video | 7 | 11 FAL.ai image models; video is an opt-in toolset with four backends |
| Audio | 9 | Eleven TTS providers plus official ElevenLabs Music and Sound Effects plugins |
| Design and publishing connectors | 7 | Canva, Gamma, Figma, Notion, WordPress.com, Webflow; no Google Docs connector |
| Safety before publishing | 5 | Approvals cover risky commands; posts need your own trust: untrusted setting |
| Ease of setup | 4 | Separate keys for each media service, plus model and gateway setup |
| Privacy and control | 9 | No telemetry; memory and skills stored locally; local models supported |
Average: 65 points over nine criteria, which rounds to 7.2.
How I scored: a 9 means the docs show a complete, ready path for the job; a 5 means it works only after you add your own guardrails; a 4 means real setup effort before the first useful draft.
Nothing here is a benchmark result, and no criterion is weighted above another.
Sources for the table: the skills page, the skills catalog, the tools reference, the TTS page, the security page and the FAQ.
Strength: Audio Is The Standout
No other part of the content stack is this complete.
text_to_speech supports eleven providers, with Edge TTS as the default that needs no key and ElevenLabs among the paid options.
The v0.23.0 release notes add that the agent can now search and pick ElevenLabs voices.
The official ElevenLabs Music plugin writes songs from 3 seconds to 10 minutes, vocal or instrumental, and can split stems or score a video.
The official Sound Effects plugin makes clips from 0.5 to 30 seconds, with an optional seamless loop.
For a podcast coach, that covers the intro sting, the music bed and an audio version of each show note.
Strength: Your Voice Becomes A File You Own
Most writing tools keep your style in a settings box.
Hermes keeps it in a skill file in ~/.hermes/skills/, built by /learn from a PDF, a docs URL, pasted notes or a workflow you just showed it.
You can read it, edit it, version it and copy it to a second profile.
A bookshop could keep one skill for staff picks and another for event write-ups, each with its own tone.
On top of skills, a global SOUL.md file sets the agent's identity in every session, so the core tone holds even when no skill loads.
Strength: Repurposing Runs On A Schedule
The cron system takes plain-language schedules, attaches skills and delivers to a chat, local files or a platform target.
That turns "make a thread from this week's video" into a job that fires every Thursday.
A per-job model pin means the long-form draft can run on a stronger model than the caption job.
Research feeds the same job.
web_search returns five results by default and accepts a limit up to 100, while web_extract returns clean page text and handles PDF links, up to five URLs per call.
The optional x_search tool reads public X posts and threads, is off by default, needs xAI credentials and never posts or replies.
A marketing freelancer could run a Monday job that gathers this week's questions in a client's niche, then a Thursday job that drafts the article and the thread from them.
The research score stops at 8 because the agent can still misread a page, so a person checks every quoted figure.
🔥 Want the setup that fixes the weak scores in this review? Inside the AI Profit Boardroom you get the installable Agent OS, video trainings on skills, cron and safe publishing, four coaching calls a week and 3,400+ members sharing what works. → Join the AI Profit Boardroom
Strength: Visual Drafts Without Leaving The Chat
Images run through image_generate on FAL.ai, with eleven models to pick from in hermes tools, according to the image generation page.
Video comes from video_generate, which makes a clip from text or animates a still when you pass an image.
In both cases the model is your setting, and the agent can't swap it on its own, which keeps output consistent across a campaign of drafts.
For layouts, the bundled skills include claude-design for one-off HTML pages and decks, and baoyu-infographic for infographics.
For designs your team already edits, Canva, Gamma and Figma sit in the Nous-approved MCP catalog.
And the official Meshy plugin turns text or product photos into 3D models, routing each credit-spending call through the approval prompt with an estimated cost.
The score stops at 7 because video needs extra setup and there's no single built-in editor for finishing a clip.
Strength: Privacy And Model Choice
The FAQ states that API calls go only to the model provider you configure and that no telemetry is collected.
It also lists local models through Ollama, vLLM, llama.cpp and other OpenAI-compatible servers.
A dental practice writing patient-education posts can keep its drafts and notes on its own machine that way.
Media plugins are the exception: ElevenLabs, FAL.ai and Meshy receive whatever prompt or file you send them.
Limit: Nothing Holds A Post Before It Goes Out
This is the biggest deduction.
Hermes approvals cover dangerous shell commands.
For MCP servers, the MCP config reference says a server marked trust: untrusted asks before every write-capable tool, and full is the default.
The catalog shows how much that varies by plugin.
- AdaptlyPost (community) marks its server untrusted, so Hermes asks before write tools and refuses them in unattended runs.
- Postel (community) is marked untrusted in its package, but its sign-in entry replaces that, so after sign-in Hermes does not ask before publishing unless you add the setting back.
- WordPress.com (Nous-approved MCP) follows the default trust level unless you change it.
So every content job needs "draft only" written into it, and every publishing server needs trust: untrusted.
Limit: Setup Is A Project
Images need a FAL.ai key, music and effects need an ElevenLabs key, and 3D needs a Meshy key plus Node.js 18 or newer.
Video is off until you enable the video_gen toolset and pick a backend: xAI, FAL.ai, OpenRouter or DeepInfra.
The docs do mention a paid Nous Portal subscription that bundles search, image generation, TTS and browser tools without separate keys, which trims some of this.
It's still more wiring than a hosted writing app.
Add a messaging gateway for review, a sender allowlist and a profile per brand, and the first afternoon goes on setup rather than writing.
The upside is that each step is documented, so nothing here needs guesswork.
Limit: Gaps In The Connector List
There's no Google Docs connector; Docs work only through the bundled google-workspace skill.
The v0.23.0 catalog has no Substack, Beehiiv, Kit, Mailchimp or Ghost entry, and no Buffer or Hootsuite.
The WordPress.com entry doesn't cover self-hosted WordPress.
And X Chat, an official plugin, is a private-message gateway into X's encrypted DMs; it doesn't post.
Posting to X goes through the xurl skill or the community Postel and AdaptlyPost plugins.
Who Should Use It
- A coach or podcaster who records weekly and wants show notes, threads and audio clips drafted from each episode.
- A small agency that wants one profile and one style skill per client, with Gamma or Canva drafts for review.
- A shop selling physical products that wants 3D previews through Meshy, where each credit-spending call asks first.
- Anyone who needs drafts kept on their own hardware, with local models where possible.
Who Should Skip It
- A café owner who wants an app that works the moment it's opened, with no keys to manage.
- Teams that need auto-posting with a sign-off step built in, since Hermes leaves that to your own settings.
- Anyone editing paid client video with the community
hermes-video-editor, whose licence is noncommercial. - Newsletter-first businesses whose platform isn't in the catalog and who don't want a copy-and-paste step.
📞 Choosing between content tools for a small team? Book a free SEO strategy session and my team will look at which content actually moves your search rankings. → Get a free strategy call
How It Scores Next To Our Marketing Review
This review grades making the pieces: drafts, audio, images, decks and 3D.
The Hermes Agent for marketing verdict grades the campaign side, such as scheduling and team rooms.
The two scores land close because the same weakness, publishing safety, drags both down.
Final Verdict
Hermes is a strong self-hosted pick for audio-heavy content and voice capture, and it lets you keep your drafts close.
It loses points for the work it takes to set up and for leaving the publish brake to you.
Write "draft only" into every job, mark every publishing server untrusted, and it becomes a safe, capable content studio.
Start with the audio and repurposing jobs, where it scores highest, and add images, video and 3D only once the first job runs cleanly.
🔥 Skip the trial and error on your content setup The AI Profit Boardroom includes step-by-step trainings, the Agent OS you can install today, four weekly coaching calls and 3,400+ members to learn from. → Join the AI Profit Boardroom
Also On Our Network
- 🌐 The nine-step setup for a draft-only Hermes content studio, commands included
- 🌐 Six small-business content jobs, from salon newsletters to 3D product previews
- 🌐 Five copy-paste prompts that turn one video into a week of drafts
- 🌐 Which AI model to pin to each Hermes content job, by real bench scores
- 🌐 The complete Hermes Agent for Content Creation guide on the AI Profit Boardroom blog
FAQ: Hermes Agent For Content Creation Verdict
What is the overall rating for Hermes as a content tool?
7.2 out of 10 on nine criteria, based on my reading of the official docs. Audio and privacy score highest; setup and publishing safety score lowest.
Why does publishing safety score only 5?
Hermes approvals cover risky shell commands, and MCP servers ask before write tools only when marked trust: untrusted. You have to add that setting and a draft-only rule yourself.
Is Hermes good for making audio content?
Yes. It has text-to-speech with eleven providers and official ElevenLabs Music and Sound Effects plugins, which is the strongest part of its content stack.
Can Hermes edit video for client work?
It can generate clips with the opt-in video_gen toolset. The community hermes-video-editor plugin has a noncommercial licence, so it is not suited to paid client work.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I've built a 7-figure SEO and link building agency, Goldie Agency, with a team of 70+ SEO experts.
- I've grown a YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery".
- My Udemy courses have taught over 29,000 students.
→ Get my best AI training inside the AI Profit Boardroom
If those scores fit your team, set up one draft-only job this week, and book a free SEO strategy session to point it at topics worth ranking for, which is the fastest way to judge Hermes Agent for content creation yourself.
