AI Agents

Hermes Open WebUI Workflows: Multi-Agent Profiles That Compound

Julian Goldie · · 7 min read

Hermes Open WebUI workflows are where most users stop after install.

Big mistake.

The install gets you a chat box.

The workflows are what turn it into a productivity system.

I'm going to walk through the 5 workspaces I run, what each does, and how they compound to save me 20+ hours a week.

What Workspaces Are In Open WebUI

A workspace in Open WebUI is a custom AI agent definition.

Each workspace has:

Think of each workspace as a custom GPT — but free, self-hosted, and self-improving.

You can have unlimited workspaces.

I run 5.

🔥 Want my full Hermes Open WebUI workspace library? Inside the AI Profit Boardroom I've put up all 12 workspaces I use across content, SEO, ops, client work, and personal AI. System prompts, knowledge bases, tool configs — the full setup. Plus weekly coaching to clone any of them into your install. Click below. → Get the workspace library

Workspace 1 — Content Drafter

The most-used workspace.

Base model: Claude Sonnet 4 (highest writing quality)

System prompt: Hormozi tone, UK English, sentence per line, my CTA structures

Knowledge base: brand voice guide, my 50 best-performing blog posts, my common phrases

Tools: web search on (for stats and recent context)

What it does: drafts blog posts, social content, emails, sales pages.

Speeds up drafting from 4 hours per post to 45 minutes (mostly editing).

If you want to see the underlying brand-voice approach, my Claude Code AI SEO post covers the principles in detail.

Workspace 2 — Research Agent

Base model: Hermes (with skills tree)

System prompt: evidence-based research, citations required, reject low-credibility sources

Knowledge base: none — pulls fresh

Tools: web search on, browser automation enabled, Firecrawl integrated

What it does: competitor analysis, topic research, fact-checking, market research.

This is the workspace I send sub-agents from for parallel research tasks.

Spawns 5 sub-agents, each researches one competitor, all report back to the main session.

I broke down the multi-agent research pattern in my paperclip Hermes agent post — pairs naturally with this workspace.

Workspace 3 — Code Reviewer

Base model: Claude Sonnet (best at code)

System prompt: code review checklist, security focus, performance concerns, suggest tests

Knowledge base: my project codebases (read-only references)

Tools: code interpreter on

What it does: reviews PRs, refactors snippets, suggests improvements, catches security issues.

I drop in a PR diff or a code snippet and Claude does a code review better than most senior devs would on a Slack ping.

The knowledge base addition matters — Claude can reference my actual codebase patterns instead of generic advice.

Workspace 4 — SEO Strategist

Base model: Hermes (skills + persistent memory)

System prompt: my SEO methodology, keyword research process, content cluster patterns

Knowledge base: my keyword lists, ranking history, top performing content

Tools: web search, Google Search Console integration via MCP

What it does: keyword research, content cluster planning, competitor SEO analysis, on-page audits.

This workspace knows my entire SEO history — what worked, what didn't, which keywords I've already targeted.

It doesn't recommend keywords I'm already ranking for.

It doesn't repeat content angles I've already covered.

That context is what makes a Hermes-powered SEO workspace dramatically better than a generic AI tool.

I covered the SEO automation angle in my DeepSeek SEO post — the principles transfer cleanly into a Hermes workspace.

Workspace 5 — Personal Ops

Base model: Hermes

System prompt: my preferences, schedule, communication style, project context

Knowledge base: project briefs, contact list, recurring tasks

Tools: Telegram bridge, Calendar MCP, Email MCP

What it does: drafts replies to emails, summarises my calendar, plans my day, triages tasks.

This is the closest thing to "personal assistant AI" I've built.

It knows me well enough that drafted replies don't need editing 80% of the time.

For the personal AI angle, my Hermes ai course post walks through the daily automation set this workspace runs.

🔥 Want my Personal Ops workspace template? Inside the AI Profit Boardroom I've put up the system prompt, the knowledge base structure, and the MCP integrations I use for my Personal Ops workspace. Plus weekly coaching to adapt it to your life. Click below. → Get the Personal Ops template

How The Workspaces Compound

Each workspace is useful on its own.

Together they're transformative.

Real example from this week:

Monday morning — Personal Ops workspace gives me a daily brief and identifies a competitor that just published a viral post.

I forward to Research Agent — 5 sub-agents analyse the post, the comments, the backlinks, the related searches.

Research Agent reports — here's what's working, here's what's missing, here are 3 angles we could take.

SEO Strategist takes the angles — runs them through my keyword history, picks the best one, plans a content cluster.

Content Drafter writes the cluster — 5 articles drafted, all in my voice, all with my CTAs.

Code Reviewer audits the website code for any technical SEO issues that could affect indexing.

End to end — Monday morning to Tuesday afternoon, I have 5 articles published, indexed, and starting to rank.

Without Open WebUI workspaces — that takes 2-3 weeks of context-switching.

The workspaces remove the context switch.

That's where the compounding comes from.

Workspace Maintenance — Don't Skip This

Workspaces drift.

Every quarter, audit them:

A maintained workspace library is a force multiplier.

A neglected one becomes a junk drawer.

Treat your workspaces like code — version, review, refactor.

I covered the same maintenance principle for Hermes skills in my Hermes vs OpenClaw post — applies equally to workspaces.

When To Add A New Workspace

Three triggers:

1. You catch yourself repeating the same instructions — turn the instructions into a system prompt and bake them into a workspace.

2. You're using a different tone or style for a specific job — separate workspace.

3. You want different model behaviour for a specific task — different model = different workspace.

Don't create workspaces speculatively.

Create them when you've felt the pain.

Hermes Open WebUI Workflow FAQ

How many workspaces is too many?

Practically — 10-15 starts feeling cluttered. I'd cap at 8-10 active.

Can workspaces share knowledge bases?

Yes — same knowledge base can be attached to multiple workspaces.

Can I export workspaces?

Yes — Open WebUI has export/import. Useful for backup or sharing with a team.

Can different workspaces use different models?

Yes — that's the point. Mix and match.

Do workspaces share Hermes memory?

Yes — Hermes memory is global to the agent. Workspaces share that base, but each can layer on its own knowledge base.

Can my team use my workspaces?

Yes if they have access to your Open WebUI instance. Each user gets their own chat history.

Related Reading

Final Take

Hermes Open WebUI workflows are the difference between "AI tool I have installed" and "AI infrastructure that runs my business".

Don't stop at install.

Build 3 workspaces this week.

Build 5 by end of month.

Maintain them quarterly.

Six months in you'll have an AI productivity system that's genuinely yours — and worth more than every paid AI tool you'd otherwise stack.

🔥 Ready to build your Hermes Open WebUI workspace library? Get a FREE AI Course + Community + 1,000 AI Agents 👉 join here. Or grab the full workspace library inside the AI Profit Boardroom.

Learn how I make these videos 👉 aiprofitboardroom.com

Video notes + links to the tools 👉 skool.com/ai-profit-lab-7462

Hermes open web ui workflows compound — start the first three workspaces tonight.

Ready to Succeed With AI?

Join 2,800+ entrepreneurs inside the AI Profit Boardroom. Get proven AI workflows, daily coaching, and a community that actually helps you win.

Join The AI Profit Boardroom →

7-Day No-Questions Refund • Cancel Anytime