6 Hermes Use Cases That OpenClaw Never Had
AAI LABS
Computing/SoftwareAdvertising/MarketingSmall Business/StartupsManagementInternet Technology
Transcript
00:00:00The Hermes agent is one of the most powerful agents out there, but people don't really know
00:00:04how powerful it can get when it's actually used to optimize the processes they interact with daily.
00:00:09Now some of you who have been using OpenClaw might think it has the same use case, and yes it does,
00:00:13but Hermes has extra features built in that OpenClaw never had, and those are what make it
00:00:18so much more powerful. Before we move on to the use cases and how they'd help you, we just wanted
00:00:22to mention that the Hermes agent now has a desktop app, which is basically a wrapper around the agent
00:00:28setup. In the previous video, we were using the terminal interface, but after testing the desktop
00:00:32app, we found it's a much more interactive way to use it than the terminal UI, because the interface
00:00:38makes it way easier to monitor the agents in action and set up the configs. Also check out that previous
00:00:43video, because using clawed code with Hermes agent opens up a lot of new use cases. The biggest
00:00:48difference for us was managing multiple profiles. A profile is basically an individual Hermes agent
00:00:53with its own separate memory and skills, which keeps your different personas isolated from
00:00:58each other. On the terminal, you'd have to run a bunch of commands just to manage and switch
00:01:02between profiles. But with the desktop app, you can run several profiles in parallel and delegate
00:01:08tasks to each one at the same time. If you want to install it, you just download the installer
00:01:12for whichever operating system you're on and run it and Hermes is ready to use. Once it's installed, you can
00:01:17configure the rest of the settings from the settings pane, where you'll set up your model, tools and
00:01:22everything else we showed you how to configure in our previous video. Since we'd already set the agent up in the
00:01:27previous video, the desktop app just picked up those same configs and carried them over. The only difference
00:01:32between the TUI and the desktop app is that now you're doing all of this through the UI. You can also configure the
00:01:38skills you want to enable from the skills panel. From that same settings panel, you can manage the memory cap we
00:01:43mentioned in our previous video, which maintains memory and profile budgets in terms of character count. We just
00:01:49wanted to mention the desktop app before diving into the use case in case you'd rather work with it, but you
00:01:54can do all of this from the TUI as well. Now there's this feature that opens up a lot of ways to use the Hermes
00:02:00agent and it's called the wake agent feature. Just like your phone that's always on do not disturb, the agent stays on it too.
00:02:06But unlike you, it actually picks up when it's needed. Wake agent is basically a flag in the script that we use
00:02:11in the cron task and it actually decides while running whether the Hermes agent should actually use the LLM or not.
00:02:17If the wake agent flag is set to false, the LLM won't get called for that event, so the agent only fires when there's an actual change worth paying attention to. What this does is save you the cost of having the LLM fire on things it shouldn't be handling.
00:02:30This wasn't possible with OpenClaw, where the agent would run on every cron job even when it wasn't needed. But with this flag, the scheduled job gets a free decision maker that decides whether spending costly LLM tokens is even justified in the first place.
00:02:42You can use it for a lot of things. For example, you can set up a group of cron jobs with skills that monitor your AWS costs along with the Gemini API you've configured for a project and if those costs rise higher than usual, it can report that in a configured channel.
00:02:55If you don't want it to just report, you can connect the Stripe MCP and other tools and let it take actions based on that too. The reason this works so well is you're not invoking the agent and wasting tokens on events where there's no cost spike.
00:03:07Instead, the agent only gets invoked when a cost spike actually happens.
00:03:11You can use the same flag to let the agent manage reviews on your Google Play Store app and for that, you just need to configure the Google Play API to access the app's reviews and set up the right cron job for it.
00:03:22But it doesn't just reply to customers leaving one-star reviews. The agent can also report those reviews in Slack and help identify the issues people are facing, sending meaningful updates so you get a proper understanding of how real users actually feel about the app and you can make improvements based on that feedback.
00:03:38The previous automation used the Wake Agent feature to call the LLM only when something needed attention.
00:03:44But there's another flag worth knowing about, called No Agent.
00:03:47This one gets the job done without ever invoking the AI model at all.
00:03:51So you might be thinking that if there's no AI model involved, then this is just a normal cron job.
00:03:55Now that is true, but what sets it apart is that it still runs inside the Hermes ecosystem.
00:04:00It works almost like a regular cron job, except that Hermes already has way more context about how you work.
00:04:06So setting one up is easier than setting up the same thing on another system.
00:04:10For example, we set up a cron job with this flag to monitor certain metrics such as the TLS to make sure the website doesn't suddenly stop working.
00:04:18It also monitors Stripe app health and other important metrics for an application and then posts the health check updates straight into a Slack channel.
00:04:25Hermes builds that cron job for you using the no agent flag set as true.
00:04:29The big upside of these jobs is they're practically free since there's no agent involved at all.
00:04:34Once you give Hermes a prompt to set one up, it pulls in the skills it needs and builds the whole automation for you, which makes it way easier to keep an eye on things day to day.
00:04:43And whenever you want Hermes to actually act on one of those alerts, whether that's fixing the issue or anything else, you just tag it in the chat and tell it what you need.
00:04:50And it takes care of everything while keeping all the existing context.
00:04:54Now, another use case we're going to talk about is implementing the Hermes agent as the second brain of your organization.
00:05:00The way it works is you connect the Hermes bot to your Slack workspace and have all your team members interact with it directly.
00:05:07This works way better than other platforms because of the evolving memory and skills it has, which let it observe your day-to-day tasks and turn them into reusable workflows while gathering so much context about you that it ends up knowing you better than your own mom does.
00:05:20So, as people use and invoke the Hermes agent, it gathers context on what the company is, what it does, all the tasks, and all the progress being made.
00:05:28Then it creates a skill that holds everything it needs to understand the company's operations, and with that skill, it acts as a complete second brain for your company.
00:05:36This way, Hermes becomes a central hub that tracks everything.
00:05:39It shares all that context across the team, so your team members can invoke it directly and ask questions or check on things across the organization.
00:05:47Your project manager can even use it to track goals because it already has the context and progress on the different initiatives, and from there, it can assign tasks to teammates and update them based on the goals you've set.
00:05:57Hermes holds up as a second brain because it does well on long-running tasks.
00:06:01A good example is each team member interacting with the agent over time, and its memory editing feature is what lets it sustain that kind of workflow.
00:06:08But an agent like OpenClaw can't keep this up because its SOL file gets so bloated that it starts losing sight of what it actually needs to focus on, so you'd have to reset that SOL file to make sure that it remains effective in the long run.
00:06:21If you like these use cases, subscribe to the channel and click the notification bell.
00:06:24On the channel, we post content that helps you learn new ways to optimize different processes in different businesses with AI.
00:06:31Your support would mean a lot to us.
00:06:33Aside from acting as your second brain, Hermes' ever-evolving memory and skills also let it take actions on your behalf.
00:06:38If you're feeling lazy, you can let it auto-respond to all your potential leads based on the info it's collected throughout your use of the agent and the skills it's built along the way.
00:06:48Hermes comes with built-in skills for accessing Google Workspace through scripts that let the agent interact with Google products.
00:06:54To connect Hermes to your Gmail, you first need to configure a project on Google Cloud and get its credentials.
00:07:00That setup is pretty long, but it's just simple steps.
00:07:03You can ask Claude or grab the setup from our community in AI Labs Pro.
00:07:07Once you've added the credentials and the callback token, Hermes is connected to Gmail.
00:07:12This unlocks everything for reading, sending, and replying to emails.
00:07:15And from there, you can use this setup for your email automations.
00:07:19With Gmail setup, the next step is to automate your processes.
00:07:22For that, you can create a webhook.
00:07:23A webhook is basically a way for one app to notify another that some event happened, so the other app can take action based on it.
00:07:30You can use Hermes to create one that replies to all the emails that look like potential leads coming into the Gmail account you've connected.
00:07:37It generates those replies based on everything it knows about the company.
00:07:40So instead of you manually sending emails, it handles the whole thing on its own.
00:07:44To do this, it uses its skills for creating webhooks along with the company skill it maintains for organizational context
00:07:50and builds a webhook that monitors incoming emails and analyzes them to spot potential customers and generate responses for them.
00:07:57So once this is in place, whenever it detects an email that looks like it's from a potential customer,
00:08:02it reaches out and responds using the company's information.
00:08:06It generates the right responses based on the context it has.
00:08:09And if a meeting is involved, it can even schedule it on your calendar based on your open time slots.
00:08:14As we talked about in a previous video, you can create a PRD as a skill.
00:08:18We covered why keeping it as a skill instead of a plain document matters.
00:08:22Because a skill only gets loaded when it's actually needed and stays in the fresh part of the context
00:08:27window where the model is paying attention so the agent never loses track of the requirements that matter.
00:08:32And since that skill already knows everything about the product you're building, you can reuse the same setup and ask it to analyze your competitors too.
00:08:39On your Hermes setup, you can create a cron job that studies those competitors and updates the PRD with potential features you could add,
00:08:46while also maintaining a separate document that tracks how those competitors are progressing.
00:08:50This saves you time, so instead of manually checking X for updates, you'll finally get more time to waste doom scrolling.
00:08:56Hermes sets up the cron job for you and the frequency is usually weekly since that's roughly how often competitors ship updates.
00:09:03The job maintains a list of competitors, creates competitive analysis files and keeps updating the PRD on its own.
00:09:10Those files basically become part of the agent's context so you can take the recommendations it generates and act on them as you're building the product.
00:09:17This way you stay on top of what your competitors are doing and have a real shot at staying ahead of them.
00:09:22And in the same way, you can build plenty of other workflows on top of the skills and context that Hermes keeps building up.
00:09:29Another thing you can hand off to Hermes is your social media.
00:09:31There's a skill called XURL which basically automates content posting to X.
00:09:36This is especially useful for us because we often need to repurpose content from one format into another.
00:09:41A good example is how we turn our long format video scripts into X posts and LinkedIn posts.
00:09:46With the Hermes agent, this becomes way easier because it already knows what our voice sounds like and how we usually write our posts since it has all that context.
00:09:55So we can create a new skill where we hand it a video script and it generates posts for both X and LinkedIn and writes them into a specific folder.
00:10:02The reason we write them into a folder instead of posting them directly is we still want a human review step in between since there are often corrections to make.
00:10:10This way we can make sure the content going to the accounts is accurate and tailored to what we want.
00:10:14Once the skill is configured, you load it with the slash command and the skill name and then hand it the video script you want posts for.
00:10:21It drafts both posts in that same folder, makes sure the X post respects the character limit and waits for us to review them so we can tailor them to our needs.
00:10:29Once we've reviewed the generated posts, we can just go to the Hermes agent and ask it to post them on X and it invokes the skill and publishes the post for us.
00:10:37But before we end the video, let's have a word by our sponsor.
00:10:40Making content about AI tools.
00:10:42I've realized using them is easy but actually understanding how they work gives you a real edge.
00:10:46That's why I've been using Brilliant and they're sponsoring today's video.
00:10:50It's an interactive platform where you build real skills in math, coding and AI by solving problems step by step instead of watching lectures.
00:10:58The courses are built by teachers from MIT, Harvard and Stanford and adapt to your pace so you're always challenged at the right level.
00:11:05Their How AI works course is one I keep coming back to.
00:11:09It breaks down neural networks, training data and decision boundaries through interactive exercises that make these concepts click like reading articles never could.
00:11:17Whether you want to strengthen your math or get more confident with coding, Brilliant works for anyone of any age.
00:11:23Our viewers get 20% off for annual premium subscription for unlimited daily access.
00:11:29To try it free for 30 days, go to brilliant.org/theailabs or click the link in the description.
00:11:34That brings us to the end of this video.
00:11:36If you'd like to support the channel and help us keep making videos like this, you can do so by using the super thanks button below.
00:11:42As always, thank you for watching and I'll see you in the next one.
Community Posts
No posts yet. Be the first to write about this video!
Write about this video