agentos.guide › blog

Hermes Agent SuperGrok: Plug Your Grok Subscription Into Hermes In Your Agent OS, With 5 Copy-Paste Prompts

By Julian Goldie · 31 October 2026 · agentos.guide

Hermes Agent SuperGrok lets the Hermes agent inside your Agent OS run on Grok using your SuperGrok or X Premium+ login, with no API key to paste.

The trick is that you don't type the setup yourself.

You hand your agent a prompt, it runs the commands, and it reports back.

Below are the quick start, five copy-paste prompts and a daily workflow built on one login.

If you'd like my team to plan the workflows with you, book a free SEO strategy session.

The Quick Start In Four Commands

These are the commands from the Hermes xAI Grok OAuth guide, in the order your agent will run them.

hermes model
hermes auth add xai-oauth --no-browser
hermes doctor
hermes tools

The first opens the provider picker, where the entry is xAI Grok OAuth (SuperGrok / X Premium+).

The second is the headless login, which prints a link and a code you approve in any browser.

The third confirms the login under ◆ Auth Providers.

The fourth sets which xAI tools the agent can use.

Once you approve, the tokens sit in ~/.hermes/auth.json and refresh on their own.

Prefer the menus?

Plain hermes model walks you through the same login and then shows the model list.

Either way, you only approve once per machine.

Prompt 1: Have Your Agent Connect Grok

Paste this into your Agent OS chat.

Connect Hermes to my SuperGrok subscription using the xai-oauth provider.
Run: hermes auth add xai-oauth --no-browser
Show me the verification URL and the user code, then wait while I approve it.
When it's approved, set the model:
  hermes config set model.provider xai-oauth
  hermes config set model.default grok-4.6
Then show me the model block from ~/.hermes/config.yaml.

The config block it shows you should read default: grok-4.6, provider: xai-oauth and base_url: https://api.x.ai/v1.

Approve the code quickly, because the guide says xAI expires it after a set window.

Prompt 2: Check The Login Actually Works

A successful login and a working model aren't the same thing.

This prompt tests both.

Run hermes doctor and tell me the state of xai-oauth under Auth Providers.
Then send one short test message through Grok and show me the reply.
If the reply is an HTTP 403, stop and tell me. Don't retry the login.

That last line matters.

The guide says a 403 after login isn't a stale token, so a fresh login won't fix it.

🔥 Want the Agent OS with Hermes and Grok already wired in?

Inside the AI Profit Boardroom you get step-by-step video trainings, the installable Agent OS and four coaching calls a week with 3,400+ members.

→ Join the AI Profit Boardroom

Prompt 3: Switch On Images, Video And X Search

The same token covers more than chat, according to the guide.

Let your agent set the backends.

Open hermes tools and set these backends:
- Image Generation: xAI Grok Imagine (image)
- Video Generation: xAI Grok Imagine (turn the toolset on, it's off by default)
- X (Twitter) Search: xAI Grok OAuth (SuperGrok / X Premium+)
Then list which xAI tools are now active.

Video needs that explicit switch.

Without it, the guide warns the agent may fall back to the bundled ComfyUI skill.

X Search is the opposite: it switches itself on once xAI credentials exist.

Prompt 4: The 403 Rescue

If Prompt 2 came back with a 403, this is the documented fix.

The issue #26847 thread shows standard SuperGrok subscribers being refused by xAI's backend after a clean login.

Grok is returning HTTP 403 after a successful OAuth login.
Move Hermes to the xAI API-key provider:
1. I'll give you my xAI API key. Add it to ~/.hermes/.env as XAI_API_KEY.
2. Run: hermes config set model.provider xai
3. Send one test message and show me the reply.
Don't print the key back to me.

The Hermes providers page lists this route exactly: XAI_API_KEY in ~/.hermes/.env, provider xai.

Per the OAuth guide, the API-key surface isn't gated the same way today.

Prompt 5: Clean Up After A Machine Or Teammate Changes

Good hygiene for any shared Agent OS.

List my stored xAI logins with: hermes auth list xai-oauth
Tell me how many entries there are and their labels.
If I say "remove all", run: hermes auth logout xai-oauth
Then run hermes doctor and confirm xai-oauth is signed out.

Logging out clears both the main entry in auth.json and any credential-pool rows.

If you only want to drop one person's login, ask the agent to use hermes auth remove xai-oauth with that entry's label instead.

Everyone else stays signed in.

Run this prompt whenever a laptop is replaced or someone leaves the team.

Prompt 6: Switch Models For A Coding Job

The picker holds more than the default, so let your agent swap models per task.

For this task only, use the coding model.
Run: hermes config set model.default grok-build-0.1
Do the task, then switch back with:
  hermes config set model.default grok-4.6
Tell me which model handled each step.

The OAuth guide lists grok-build-0.1 as the coding-oriented Grok Build model.

For long documents, grok-4.3 is the one to try, because the v0.14.0 release notes gave it a 1M token context window.

Keep grok-4.6 as your everyday default, since it's the one pinned at the top.

A Daily Workflow On One Login

Here's a simple day that runs entirely on the SuperGrok login.

Each step uses a model or tool from the OAuth guide's own tables.

Schedule the morning and end-of-day steps, and trigger the middle two by hand.

The end-of-day check is the safety net.

If the login has gone stale, you hear about it that evening, not halfway through tomorrow's work.

If a refresh fails for good, Hermes quarantines the dead token and shows a single re-authentication message, so your agent just runs hermes auth add xai-oauth again.

Sharing Grok With Your Other Coding Tools

Your Agent OS may already run Codex CLI, Aider, Cline or Continue.

The Hermes v0.14.0 release notes introduced hermes proxy, which serves an OpenAI-style API on localhost using your OAuth login.

SuperGrok is one of the providers named.

v0.15.0 then gave the proxy an xAI upstream.

Ask your agent to start hermes proxy and point one tool at the localhost address it prints.

🔥 Want my team to plan your Agent OS workflows?

Book a free SEO strategy session and we'll map which jobs your agent should take on first, and which model should run each one.

→ Book your free call with my team

Why Let The Agent Do The Setup

Typing commands yourself works, but it doesn't scale.

Once a prompt works, you can reuse it on every new machine in your Agent OS.

Your agent also reads the output for you, so a 403 or a timeout gets spotted straight away.

And you never have to remember which of the two xAI providers you picked last time.

Save each prompt above as a skill or a note, and the whole setup becomes a two-minute job.

Which Grok Model Does What

The OAuth guide's model table gives your agent plenty to choose from.

For chat, the options are grok-4.6, grok-build-0.1, grok-4.3 and three grok-4.20-0309 variants for reasoning, non-reasoning and multi-agent work.

For images, there's grok-imagine-image at roughly 5 to 10 seconds, plus grok-imagine-image-2.0 and grok-imagine-image-quality at roughly 10 to 20 seconds.

For video, there's grok-imagine-video for text-to-video and grok-imagine-video-1.5-preview for image-to-video.

Tell your agent which job it's doing, and let it pick from that list.

Hermes also turns on prompt caching automatically for any x.ai address, which speeds up long conversations without any setting.

What To Keep An Eye On

Three things, all from the docs.

• The model list refreshes from the models.dev cache, so new Grok models appear on their own, with grok-4.6 pinned on top.
• xAI retired several older models on 15 May 2026, and hermes migrate xai rewrites old configs, as a dry run unless you add --apply.
• The Hermes TTS page lists xAI TTS with XAI_API_KEY, so add the key if voice fails on the login alone.

The provider itself arrived in Hermes v0.14.0 on 16 May 2026.

Want the full Agent OS with these prompts saved and ready? Join the AI Profit Boardroom.

FAQ

Can my agent set up the SuperGrok login itself?

Yes, apart from one step. The agent runs the commands and shows you the verification link, and you approve it in your browser.

Why tell the agent not to retry the login on a 403?

Because the guide says a 403 after login isn't a stale token. Re-running the login changes nothing, so the agent should move to the API-key fix instead.

Where does my agent put the xAI API key for the 403 fix?

In ~/.hermes/.env as XAI_API_KEY, then it switches the provider to xai with hermes config set model.provider xai.

Does the Agent OS need a tunnel for a remote login?

No. The headless command prints a link and a code you can approve from any browser, and the OAuth guide says xAI needs no SSH tunnel.

Final word on Hermes Agent SuperGrok

Paste Prompt 1 into your agent now, and if you want the whole system ready to go, join the AI Profit Boardroom and run Hermes Agent SuperGrok inside the Agent OS today.

Related guides on agentos.guide

→ Hermes Camofox: Run The Stealth Browser In Your Agent OS

→ Best Hermes Agent Model: The 5-Lane Routing I Run

→ Grok Bot vs Codex: Run Both As One Crew In Your Agent OS

→ New Grok Bot vs Hermes Agent — The Ghost Team

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

← all guides on agentos.guide