01What Hermes Agent Is and How It Works
Hermes agent is a free open-source AI that can build and automate anything for you. It works like a 24/7 employee that remembers everything about you and your business. Instead of waiting for you to tell it what to do, it can go off and do tasks on its own.
Hermes agent uses a terminal interface on your computer. You give it a task in plain English, and it writes code, installs software, or runs automations. It can connect to free APIs or local models, so you don't need a paid subscription to start.
In this episode, Julian Goldie shows how to set up Hermes agent, give it a powerful memory, and schedule tasks so it works autonomously.
02Main Features of Hermes Agent
Hermes agent has several key features that make it useful:
- Open source and free – You can download it from GitHub and run it without paying anything.
- Autonomous scheduled tasks – Ask it to do something every day at a specific time, like research AI news at 9 AM.
- Memory system – It can remember your projects, goals, and writing style when connected to Obsidian and OMI.
- Team agent support via Paperclip – Paperclip lets you create a full team of specialized agents with a dashboard.
- Free API support – Use models like DeepSeek v4 Flash, Step 3.5 Flash, or our alpha on OpenRouter at no cost.
- MCP (Model Context Protocol) for Claude – You can connect Hermes agent to Claude so Claude can control Hermes.
03How to Set Up Hermes Agent Step by Step
Follow these steps to get Hermes agent running on your computer for free:
- Install Hermes agent – Go to the official GitHub page, copy the quick start instructions, and paste them into your terminal.
- Choose a free model – After installation, open the Hermes setup menu, select your provider (like News Portal), and pick a free model such as DeepSeek v4 Flash or Step 3.5 Flash.
- Set up Paperclip (optional) – If you want a team dashboard, copy the GitHub URLs for Paperclip and its Hermes connector, paste them into Hermes, and ask it to install them.
- Give your agent a memory – Install OMI (records your screen and microphone) and Obsidian (stores notes). Export OMI notes to Obsidian daily.
- Organize your memory vault – Ask Hermes or Claude to organize your Obsidian vault into sections like projects, resources, and archive with a readme file.
- Schedule your first task – Type a command like “Create a scheduled task that generates five proactive ideas each day to improve my business.”
04How to Use Hermes Agent: Examples from the Episode
Julian walks through several ways to use Hermes agent effectively:
Build a team with Paperclip – After installing Paperclip, log into its dashboard, define a goal (e.g., “grow the airport for the board community”), and set up agents for each role (engineer, tester, CEO). Assign tasks by creating new issues.
Set up scheduled tasks for autonomy – If you feel like you always have to prompt your agent, tell it to schedule a task. For example, say “Create a scheduled task that generates five proactive ideas each day to help me automate more.” The agent will then work daily without you.
Connect Hermes to Claude via MCP – Open a Claude chat, paste the Hermes MCP setup instructions from GitHub, and ask Claude to set it up. It will handle the installation and give you confirmation.
Fix a buggy extension – If Hermes Workspace isn't syncing, paste its GitHub details into Hermes agent and ask it to fix the issue. You can also ask Claude to test and fix it in parallel.
05Results, Tests, and Comparisons Shown in the Episode
Julian compares different approaches to running Hermes agent:
- Free APIs vs. local models – Free cloud APIs like DeepSeek v4 Flash or Step 3.5 Flash give much better results than local models. Even on a Mac Studio, local models are slow and lower quality. For lightweight needs, Gemma 4 (designed for mobile) works on almost any setup.
- OMI + Obsidian memory – Julian shows his own organized vault with a knowledge graph, index, daily templates, and sections for each project. He proves that raw OMI exports (1,200 memories in one document) become useless unless organized by an AI like Claude or Hermes.
- Scheduled tasks vs. constant prompting – He demonstrates that setting a scheduled task (“research latest AI automation news every day at 9 AM”) eliminates the need to keep asking the agent to work.
06Tips, Limits, and Mistakes to Avoid
Julian shares several important warnings and tips:
- Free APIs log your data – If you use free APIs from OpenRouter or News Portal, your prompts and completions may be logged by the provider. Do not send private or sensitive information through them.
- Hermes Workspace can be buggy – This free extension sometimes doesn't sync properly. If you have issues, paste the GitHub details into Hermes agent and ask it to fix the problem. You can also ask Claude for help.
- Don't overcomplicate with multiple agents – For 99% of people, you don't need an orchestrator or multiple specialized profiles. One Hermes agent with scheduled tasks works well enough.
- Don't build agents upside down – Starting with an orchestrator and lots of context before creating specialized agents often leads to frustration. Simplify: use one agent, schedule tasks, and add memory.
- Avoid token drain – Use free models or run local lightweight models like Gemma 4 to keep costs at zero.
07Who Hermes Agent Is For and Why It Matters
Hermes agent is for anyone who wants a free way to automate repetitive tasks without needing to code. It's especially useful for business owners, freelancers, and content creators who want to generate leads, build websites, or manage email outreach.
Julian explains that clients care about results – more customers and leads – not the automation itself. So if you offer Hermes setups as a service, you can charge for one-click installations, consulting calls, or monthly retainers for maintenance.
The biggest value of Hermes agent is that it can work autonomously once you set up scheduled tasks. It stops waiting for you and starts delivering daily. Combined with a good memory system (OMI + Obsidian), it becomes a true 24/7 AI employee.
