Hermes Agent Goals: Free Setup for Autonomous AI Workflows

01What Hermes Agent Goals Is and How It Works

Hermes agent goals is a free update that lets you give the AI a long-term objective and have it keep working until the goal is met, without you needing to type 'keep going' over and over. You just set a target, walk away, and come back to finished work.

After the agent completes a turn, a lightweight judge model reads the last response and decides if the goal is done or if the agent should continue. If the goal is not met, Hermes automatically creates a continuation prompt and feeds it back into the same session. This loop keeps running until the turn budget runs out or the judge confirms the goal is complete.

The goal stays in context every turn, so the agent never forgets the original target. You can close your laptop, go to sleep, and resume later with the /resume command.

02Main Features of Hermes Agent Goals Update

Hermes agent goals brings several key features that make it a true autonomous AI employee. First, the judge model runs after every turn to check if the goal is done. It is deliberately conservative — it only marks the goal done when the response explicitly confirms completion or the deliverable is clearly produced.

• Turn budget: Default is 20 turns, but you can change it. The agent auto-pauses at the limit, and you can extend the budget.
• Continuation loop: Hermes automatically appends a plan user role message to the history, keeping context alive across turns.
• Persistence: Goal state saves locally. You can close your laptop, come back tomorrow, and run /resume to continue where you left off.
• Sub-goals: You can set sub-goals with text to tighten success criteria mid-loop without breaking the whole workflow.

All these features work the same on CLI, Telegram, Discord, or an agent operating system.

03How to Set Up Hermes Agent Goals for Free

Setting up Hermes agent goals is completely free because Hermes agent is open-source. The goal skill is baked directly into Hermes, so you do not need extra tools or subscriptions. Follow these steps to get started:

1. Install Hermes agent using the official open-source package. Instructions are on the Hermes GitHub page.
2. Choose a gateway — Hermes works on CLI, Telegram, Discord, or you can use an agent operating system like the one Julian shows.
3. Set the API: Hermes can use free APIs like Nous Alpha on OpenRouter, which is free. You can also use other LLMs such as Grok 4.3 or Claude.
4. Type the /goal command followed by your objective. For example: '/goal Create a blog about OpenClaw, make it hosted so I can preview it, add a few blog posts, and make it look beautiful.'
5. Hit launch goal — the agent starts immediately. You can now close your laptop and come back later.

You do not need to watch every step. The system runs in its own directory, so it never interferes with your other work.

04How to Use Hermes Agent Goals: Step-by-Step Walkthrough

In the episode, Julian shows how to use Hermes agent goals with a real example. First, he opens Hermes and types a goal: 'go off, create a blog about OpenClaw and make sure it's hosted so I can preview it, add a few blog posts, plus make it look beautiful.' He then clicks 'launch goal,' and the system starts running autonomously.

The judge model checks after each turn. If the blog is not fully built, the continuation loop sends a new prompt to keep working. Julian shows the agent still running in the background while he explains the features.

Later, he demonstrates a previous completed goal — a simple haiku about a lobster — to show how goals work end-to-end.

You can use the following commands with Hermes agent goals:

• /goal — starts a new goal.
• /goal status — shows the current status, number of turns used, and sub-goals.
• /goal pause — pauses the goal mid-loop.
• /goal resume — resumes a paused goal.
• /goal clear — clears the current goal.

This walkthrough shows that you can automate big tasks like creating websites without being stuck at your computer.

05Results, Tests, and Comparison: Old Way vs. Hermes Goals

Julian compares the old way of using AI agents with the new Hermes agent goals system. With the old way, you type a goal, the agent does one turn, then stops. You then type 'keep going' again and again, repeating this 10 or 20 times.

Each prompt loses context because the agent forgets the original goal halfway through. If you walk away for an hour, the agent sits idle waiting for your next 'keep going.' The result is that agents only do small tasks, and you are the bottleneck.

With Hermes goals, you give one /goal command, and the agent starts immediately. The judge model decides if it is done or continues. There is zero human attention needed mid-task.

The original goal text stays in context every turn, so the agent never forgets. You can close your laptop, come back tomorrow, and resume — the goal persists. The result is long-horizon work that finishes on its own.

Julian notes that this is similar to Codex's /goals feature but adapted to Hermes architecture. He mentions that the judge is correct about 90% of the time, and you can trust it for most tasks.

06Tips, Limits, and Mistakes to Avoid with Hermes Goals

When using Hermes agent goals, there are a few things to keep in mind. First, the judge model might get it wrong sometimes, but Julian says it works correctly about 90% of the time. Do not let the fear of occasional mistakes stop you from using the system.

• Do not watch every step: Each goal runs in its own directory, so nothing collides with your work. You are safe to walk away.
• Do not keep typing 'keep going': That makes you the bottleneck. Use /goal instead and let the agent run autonomously.
• Adjust the turn budget: The default is 20 turns, but you can change it. If you know a task needs more steps, increase the budget. If you want to resume in chunks, the auto-pause at the limit helps.
• Use sub-goals wisely: You can add a sub-goal mid-loop to tighten success criteria without breaking the current process.
• Persistence: Goal state saves locally, so you can close your laptop and come back tomorrow with /resume. Do not forget to run /resume to continue.

Avoid the limiting belief that you need to watch every step. Trust the judge, set the budget, and let Hermes work.

07Who Hermes Agent Goals Is For and Why It Matters

Hermes agent goals is for anyone who uses AI agents and wants to automate large tasks without sitting at the computer. It matters because it removes you as the bottleneck — you set a target, walk away, and come back to finished work. This means you can create long-horizon projects like building a full website, writing multiple blog posts, or automating research.

It is especially useful for developers, content creators, and business owners who want to get more done in a day than most people get done in a week. The system is free and open-source, so there is no barrier to entry. It works with free APIs like Nous Alpha on OpenRouter.

Julian recommends using Hermes goals together with tools like Claude, OpenClaw, Codex, Anti-Gravity, Gemini, and other AI agents in a unified operating system. If you want a 30-day roadmap and weekly coaching calls, you can join the AI Profit Boardroom. But even without that, the core feature is free and ready to use.