Hermes Agent Image Generation - Free Setup Guide (2026)

Julian Goldie — founder, AI Profit Boardroom
By Julian Goldie · 7 min read
Get The AI Profit Stack Join AIPB →
🎯 1,000+ done-for-you AI agent workflows 📅 5 live coaching calls / week with me 🛡️ 7-day refund + 30-day ROI guarantee 👥 3,000+ AI operators inside

So hermes agent image generation is finally here: your Hermes agent can now generate images using ChatGPT Images 2.0 — the GPT-Image-2 model — and there is a route that costs absolutely nothing. The Hermes co-founder announced you can use OpenAI auth to generate images with Hermes, and Julian Goldie put it to the test on camera, errors and all.

This guide covers his exact setup flow, the free Codex route, the more reliable API key route, and the honest troubleshooting that got it all working.

📺 Watch: Hermes AI: How to Setup ChatGPT Image 2.0!

What Is Hermes Agent Image Generation?

Hermes is the free AI agent platform Julian Goldie runs his automation through, and image generation is the newest tool in its kit. Once it is configured, you ask your agent for an image in plain English and it comes back with a finished picture, powered by ChatGPT Images 2.0.

That matters because images are one of the most repetitive jobs in any content operation. Inside Julian Goldie's Agent OS, image generation is one of the core skills his agents lean on every single day — YouTube thumbnails, blog post images, visuals for guides — so plugging a capable image model directly into Hermes removes an entire separate tool from the stack. The agent that writes the content can now illustrate it too.

In practice, the announcement gives you two routes: a completely free one through Codex, and a paid one through an OpenAI API key that behaves far more consistently.

Codex Auth vs OpenAI API Key: Your Two Options

Before you pick a lane, it helps to see both routes side by side. Julian Goldie ran both through his usual Goldie Bench style of testing — hands-on, on camera, with the failures left in — so the reliability notes below are his real-world findings from one machine, not official claims.

FactorCodex Auth (free)OpenAI API key
CostNothing — it piggybacks on your existing ChatGPT session via Codex, and Codex has a free tierPaid — standard OpenAI API usage, with a key from openai.com
ModelGPT-Image-2 (ChatGPT Images 2.0)GPT-Image-2 (ChatGPT Images 2.0)
SetupPress spacebar to select Codex Auth — Hermes says no configuration is needed, and no API key is involvedGrab a key from openai.com and plug it into the Image Generation tool
Reliability in his testingA bit buggy — sometimes it works first time, sometimes it needs a nudgeMuch smoother — images generate cleanly and save locally
Best forAnyone who wants a completely free image pipeline and does not mind occasional frictionAnyone who depends on images daily and values reliability over cost

📺 Watch: Hermes Agent V0.19 Just Changed AI Agents Forever!

How to Set Up Hermes Agent Image Generation (Step by Step)

Here is the exact flow Julian Goldie followed in the video — four steps, all in the terminal.

  1. Update Hermes. Run hermes update in your terminal so your install picks up the new image generation feature.
  2. Open the tools menu. Run hermes tools, find the Image Generation tool in the list, and choose to reconfigure it.
  3. Select Codex Auth (free). Press spacebar to select it. Hermes confirms there is no configuration needed — it uses GPT-Image-2 via Codex with no API key at all.
  4. Restart. Exit the session and run gateway restart so the change actually takes effect.

After the restart, ask your agent for an image. In Julian Goldie's run it took a couple of minutes for the chat to recognise the new tool, so give it a moment before you assume something is broken.

Why the Free Route Costs Nothing

It sounds too good to be true, so here is the mechanism. Hermes piggybacks on your existing ChatGPT session via Codex. Codex has a free tier, and Hermes itself is free. There is no hidden meter running anywhere in that chain, which is why the whole image pipeline costs nothing.

Troubleshooting: What Actually Went Wrong

This is the part most tutorials quietly cut out. Julian Goldie left it all in, because his first attempts did not work.

The initial image requests errored with the message "the image generation service isn't configured in this environment" — even after the update, the tool change and the restart. There was some back and forth before it behaved. Two fixes got it over the line.

Fix 1: Let an AI Repair the Install

Rather than digging through configuration by hand, he got Claude to help. He gave it the Hermes GitHub details and asked it to fix the local install. Claude worked through the problem, and after that the image generation started working — though it still took a couple of minutes for the chat to recognise the tool before the first image came through.

That trick is worth remembering: when an agent tool misbehaves, pointing another AI at the project's GitHub details is often faster than debugging it yourself.

Fix 2: Watch for Silent Tool Switching

One more gotcha from the video. At one point Hermes silently switched to a different image generator — Stable Diffusion — instead of the new one. The fix was refreshingly simple: he told the agent to use the codex image gen tool, and it did.

So if your images suddenly look different, ask the agent which tool it just used. One clear instruction pulls it back onto GPT-Image-2.

📺 Watch: Hermes Agent + Buzz is INSANE!

Option 2: The OpenAI API Key Route

If the free route sounds like more friction than you fancy, there is a smoother path. Get an API key from openai.com and plug it into the same Image Generation tool inside hermes tools.

In Julian Goldie's testing this route was much more reliable. Images generated cleanly on the first ask and saved locally, without any of the back and forth the Codex route needed. The trade-off is obvious: you are now paying standard API usage instead of riding a free tier.

📺 Watch: Hermes Agent: New Voice Activation Update is INSANE!

The Goldie Bench Verdict: Free vs Paid

Here is where his hands-on comparison landed. Keep in mind this is one person's testing on his own setup — a Goldie Bench finding, not a laboratory benchmark.

His conclusion: your choice. If you are experimenting, Codex Auth is hard to argue with at a price of nothing. If image generation sits inside a workflow you rely on daily, the API key pays for itself in saved patience.

If you want your Hermes agent generating images that make you money, check out the AI Profit Boardroom — it is where Julian Goldie shares the exact agent setups, prompts and workflows behind everything he tests on camera, so you can copy what already works instead of debugging alone. → Get the full Hermes image setup inside

Bonus: Can Claude Code Be Free Too?

A related question came up in the same video: Hermes is free, the images are now free — can Claude Code be free as well?

Yes, it can. You can run it through Ollama with local models, for example Gemma 4. It works, but in his experience it is not as good as the full setup. The pattern matches the image tools: free gets you moving, paid gets you consistency.

FAQ

Is Hermes agent image generation really free?

The Codex Auth route is, yes. Hermes is free, Codex has a free tier, and Codex Auth simply uses your existing ChatGPT session — so nothing in the pipeline carries a cost. The alternative route, an OpenAI API key, is paid.

Which model does Hermes use for images?

ChatGPT Images 2.0, running on the GPT-Image-2 model. Both routes use the same model — the difference in Julian Goldie's testing was reliability, not the model.

Why does my agent say the image service is not configured?

Julian Goldie hit exactly that error. Make sure you have run hermes update, reconfigured the Image Generation tool under hermes tools, selected Codex Auth, and fully restarted with gateway restart. If it still fails, do what he did: give an AI assistant such as Claude the Hermes GitHub details and ask it to fix the local install.

Why did my agent use Stable Diffusion instead?

It happened in the video too — Hermes silently switched to a different image generator at one point. The fix was telling the agent to use the codex image gen tool. One instruction, sorted.

What would I actually use this for?

Whatever your agents already produce. In Julian Goldie's world that means thumbnails and blog images generated by the same Agent OS workflows that write the content — the image step becomes one more skill the agent runs on its own.

The Bottom Line

Hermes agents generating images with ChatGPT Images 2.0 is a genuinely useful upgrade, and the setup takes minutes: run hermes update, open hermes tools, reconfigure Image Generation, press spacebar on Codex Auth, then exit and run gateway restart. Expect a little friction on the free route — Julian Goldie certainly hit some — and know the OpenAI API key route is there when you want reliability.

Free but occasionally moody, or paid and smooth — either way, your agent now makes its own images. That is one more job you never have to do by hand again.

Real wins from inside the AI Profit Boardroom

See all 3,000+ members →
AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot AIPB member win screenshot

Ready To Join The #1 AI Community?

Join 3,600+ entrepreneurs inside the AI Profit Boardroom. Get 1,000+ plug-and-play AI agent workflows, daily coaching, and a community that holds you accountable.

Join The AI Community →

7-Day No-Questions Refund • Cancel Anytime

← Back to all posts