Hermes Workspace Profiles: Multi-Agent Made Easy

Hermes Workspace profiles are how you actually run multiple AI agents without going crazy — here's how to use them.

If you've been running one Hermes agent for everything, this changes that.

This post covers:

What Hermes Workspace profiles are.
How they differ from Hermes Swarm.
How to set up your first profile.
5 specific profile templates to copy.

What Profiles Are

Hermes Workspace profiles let you have multiple AI agents in one setup.

For each profile:

Different system prompt.
Different model.
Different skills.
Different memory.

Switch between them with one click.

That's it.

Simple but powerful.

Profiles Vs Swarm

These get confused.

Profiles:

Multiple agents.
You switch between them manually.
Each is independent.

Swarm:

Multiple agents.
Aurora orchestrator coordinates them automatically.
They work together on missions.

Profiles = manual switching.

Swarm = automatic orchestration.

You can use both — profiles for general management, Swarm for coordinated missions.

I cover Swarm in Hermes Swarm.

Why Profiles Matter

Three reasons.

1 — Specialisation

Different work needs different agent setups.

A research agent shouldn't have the same prompt as a customer support agent.

Profiles let you specialise.

2 — Cost optimisation

Match cheap models to easy tasks.

Use expensive models only for hard tasks.

Profiles enable per-agent model choice.

3 — Mental clarity

When you switch profiles, you're switching modes.

"Now I'm doing research" vs "Now I'm doing customer ops."

Helps focus.

Setting Up Your First Profile

In Hermes Workspace settings, find Profiles section.

Step 1 — Click "New Profile"

Give it a name (e.g. "Researcher").

Step 2 — Define system prompt

What's this agent's role?

Example for Researcher:

"You are a research specialist. For any topic, find verified sources, synthesise findings, output as a structured brief."

Step 3 — Pick a model

Match to task:

Researcher → long-context cloud model.
Writer → creative cloud model.
Customer Ops → fast cheap model.
Code → code-specialised model.

Step 4 — Configure skills

Enable only skills relevant to the role.

Researcher needs web search.

Writer needs less.

Code needs terminal access.

Step 5 — Set up memory

Each profile has its own memory.

Configure what context this agent should remember.

Step 6 — Save

Profile is ready to use.

🔥 Want my full Hermes Workspace profile templates? Inside the AI Profit Boardroom, I share my exact profile templates for Researcher, Writer, Customer Ops, Code, and more. Plus 2-hour Hermes course and weekly live coaching. 2,800+ members. → Get the templates

5 Profile Templates To Copy

Real profiles I run.

Profile 1 — Researcher

System prompt: "You are a research specialist. For any topic, find verified sources from at least 5 places. Synthesise findings into a structured brief with executive summary, key findings, and citations. Don't speculate — cite sources."

Model: Long-context cloud (Claude Opus or Kim K2.5).

Skills: Web search, Tavly, browser, file ops, memory.

Use for: Market research, competitor analysis, deep dives.

Profile 2 — Writer

System prompt: "You are a content writer in Julian's voice. Hormozi tone, sentence-per-line, UK grammar. SEO-aware. Always include keyword in first/last line, FAQ section, internal links, multiple CTAs."

Model: Creative cloud (GPT or Claude).

Skills: Web search, file ops, memory.

Use for: Blog posts, social media drafts, email copy.

This is what I run for Claude Code SEO Agent workflows.

Profile 3 — Customer Ops

System prompt: "You handle inbound customer messages. For FAQs, give the answer with a relevant link. For pricing, refer to my pricing page. For support issues, gather context (email, what went wrong, what they tried) and escalate to me. Be friendly but concise."

Model: Fast cheap cloud (Gemma 4 or DeepSeek free tier).

Skills: Email, calendar, memory.

Use for: Telegram, WhatsApp, email triage.

Pairs with Telegram AI Agent.

Profile 4 — Code

System prompt: "You are a code assistant. For any code task: read existing code, understand patterns, write clean implementations, add tests where needed. Don't break existing functionality. Prefer explicit over clever."

Model: Code-specialised (Qwen 3.6 or Claude Sonnet).

Skills: Terminal, file ops, code analysis.

Use for: Bug fixes, refactors, new features.

Profile 5 — Daily Review

System prompt: "You help me review my day. Take notes I've made, summarise key wins, identify what needs follow-up, queue tomorrow's priorities."

Model: Mid-range cloud.

Skills: Memory, calendar, file ops.

Use for: End-of-day reflection.

Switching Between Profiles

In Workspace, profile switcher is in the sidebar.

Click to switch.

Active profile changes:

The system prompt applied.
The model used.
The skills available.
The memory accessed.

Quick context switch.

Combining Profiles With Swarm

Pro tip: profiles work alongside Hermes Swarm.

Swarm uses preset roles (Builder, Reviewer, etc).

Profiles are your personal agent configs.

Use Swarm for coordinated missions.

Use profiles for individual workflows.

I cover Swarm in Hermes Swarm.

Common Profile Mistakes

1. Vague system prompts.

"Help me with stuff" produces bad output.

Be specific about role and behaviour.

2. Same model for every profile.

Defeats the cost optimisation benefit.

Match model to role.

3. Too many skills enabled.

Bloats context.

Slows agent.

Enable only what each profile needs.

4. No clear use case per profile.

If you don't know when to use a profile, you won't use it.

Define use case before creating.

Profile Memory Patterns

Each profile has its own memory.

Common patterns:

Researcher profile memory

Topic preferences.
Source preferences.
Output formats you like.

Writer profile memory

Voice/tone notes.
Brand guidelines.
Common CTA patterns.

Customer Ops profile memory

Common customer questions.
Pricing details.
Escalation rules.

Memory makes profiles smarter over time.

Profile Best Practices

Three things that work.

1 — Document your profiles

Keep a note of what each profile does and when to use it.

2 — Iterate weekly

Review which profiles you actually used.

Refine prompts based on results.

3 — Don't over-create

Start with 2-3 profiles.

Add more only when you have specific use cases.

5-7 profiles is the sweet spot for most users.

How Many Profiles Is Too Many?

Past 10 profiles, management overhead rises.

For most users:

3-5 profiles for focused work.
5-7 profiles for varied work.
10+ only for complex multi-domain operations.

Daily Reality With Profiles

What it looks like:

8 AM — switch to Researcher profile, deep dive on a topic.
9 AM — switch to Writer profile, draft content.
11 AM — switch to Customer Ops, handle inbound messages.
2 PM — switch to Code profile, work on side project.
6 PM — switch to Daily Review, wrap up the day.

Five distinct contexts.

One Hermes Workspace.

🚀 Want my full profile + Swarm playbook? The AI Profit Boardroom has my Hermes Workspace profile templates, 2-hour Hermes course, OpenClaw 6-hour course, daily training, weekly live coaching. 2,800+ members. → Join here

FAQ — Hermes Workspace Profiles

How many profiles can I have?

Practically unlimited.

5-7 is the sweet spot.

Can profiles share memory?

By default, no.

You can configure cross-profile memory if needed.

Can two profiles use the same model?

Yes — model choice is per profile, but they can be the same.

Are profiles the same as Swarm sub-agents?

No — profiles are individual configs you switch manually.

Swarm sub-agents are coordinated automatically.

Can I delete a profile?

Yes — anytime.

Will my chat history per profile be saved?

Yes — each profile has its own chat history.

Can I export/import profiles?

Limited — share configs as text for now.