Vibe Marketer Academy
Guide · Local AI

Set Up a Nous Hermes Agent

A local AI agent for privacy-first marketing workflows

Nous Hermes is an open-source language model fine-tuned for instruction-following and agentic tasks. Unlike Claude or GPT-4, it runs entirely on your computer: no data sent to any server, no per-token costs, no rate limits. This guide sets up Nous Hermes as a local marketing agent for tasks where privacy or cost are constraints: processing customer data, drafting at volume, and running workflows that never leave your machine.

45 minAdvancedA running local Nous Hermes agent
Set it up →Browse guides
The right tool

Local AI is not always better. But sometimes it is the only option.

Claude is better at most marketing tasks. Nous Hermes is better in specific situations where Claude cannot be used.

Use Claude when...
Use Nous Hermes when...
You need the best possible output quality
You are processing sensitive customer data that cannot leave your machine
Speed matters and you have a budget
You need to run high-volume inference with no per-token cost
You are doing one-off or low-volume tasks
You are offline or in a restricted network environment
You need the latest knowledge and capabilities
You want to fine-tune the model on your own proprietary data
Requirements

Hardware and software checklist.

Hardware

Nous Hermes 2 (7B) runs on a Mac with 16GB RAM or a PC with an NVIDIA GPU with 8GB+ VRAM. The 13B model needs 32GB RAM or 16GB VRAM. If you are on consumer hardware, start with the 7B model.

Ollama

The easiest way to run local models. Free, open-source, one-command install. Works on Mac, Windows, and Linux.

Python 3.10+

For building the agent wrapper that lets you run marketing workflows against the local model.

Git

For cloning the agent template. Install from git-scm.com if you do not have it.

Step by step

From zero to running agent in 45 minutes.

1

5 min

Install Ollama

Ollama handles model downloading and serving locally.

# Mac
brew install ollama

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# Windows: download the installer from ollama.com
2

10 min

Download Nous Hermes 2

Pull the model. This downloads approximately 4GB for the 7B model.

# 7B model -- runs on most modern hardware
ollama pull nous-hermes2

# 13B model -- better quality, needs more RAM
ollama pull nous-hermes2:13b

# Verify it is working
ollama run nous-hermes2 "What is generative engine optimization?"
3

5 min

Set up the agent wrapper

Create a Python script that wraps Ollama with a marketing-focused system prompt and task runner.

mkdir hermes-agent && cd hermes-agent
pip install requests python-dotenv
touch agent.py
4

15 min

Write the agent

Open agent.py and paste this starter agent. It connects to your local Nous Hermes instance and runs marketing tasks.

import requests
import json

OLLAMA_URL = "http://localhost:11434/api/generate"
MODEL = "nous-hermes2"

SYSTEM_PROMPT = """You are a professional marketing assistant. You help with:
- Writing copy: ads, emails, landing pages, social posts
- Marketing strategy: positioning, messaging, channel planning
- Content: blog posts, newsletters, scripts
- Research: competitor analysis, audience research, trend summaries

You produce clear, specific, professional output. When writing copy, you always
produce multiple variants. You ask clarifying questions when the brief is incomplete.
You never use em-dashes, and you avoid generic marketing language."""

def run(task: str, context: str = "") -> str:
    prompt = f"{context}\n\nTask: {task}" if context else task
    payload = {
        "model": MODEL,
        "system": SYSTEM_PROMPT,
        "prompt": prompt,
        "stream": False
    }
    response = requests.post(OLLAMA_URL, json=payload)
    return response.json()["response"]

if __name__ == "__main__":
    # Example: write three email subject line variants
    result = run(
        task="Write 5 subject line variants for an email promoting a free AI marketing guide. The audience is small business owners aged 30-50 who are skeptical of AI hype.",
        context="Brand voice: calm, practical, no hype. Never use exclamation points."
    )
    print(result)
5

5 min

Run your first task

# Start Ollama server (if not already running)
ollama serve

# In a new terminal, run the agent
python agent.py
Use cases

What to use your local agent for.

Bulk content drafts

Generate 50 social post drafts, 20 email subject lines, or 100 ad copy variants at zero marginal cost. Run overnight, review in the morning.

Private customer data analysis

Summarize customer reviews, analyze support tickets, or extract themes from sales call notes without sending any of that data to a third-party API.

Fine-tuning on your own content

Train Nous Hermes on your past writing to create a model that sounds genuinely like you, not a generic AI assistant.

Offline-first workflows

Run marketing workflows on a plane, in a client facility with restricted internet, or anywhere API access is not available.

Community

Build alongside
other marketers.

Ask questions, share what you're building, and get unstuck faster. The community moves at the pace of people who are actually shipping.

Ask anything

No question too basic

Share your builds

Real feedback from peers

Early access

New guides first

Join the communityOr start reading first

Free to join · No credit card · Unsubscribe any time