7 Hermes Uses You Must Try (That Actually Work)
In this video, Andrew and Eric—two entrepreneurs deeply embedded in the AI agent ecosystem—explore real-world, practical applications of Hermes, a powerful AI agent designed to automate complex workflows. Andrew, a content creator and builder, demonstrates his excitement for Hermes' evolving capabilities, while Eric, who runs a company heavily reliant on AI agents, provides grounded, critical feedback based on his own extensive experience. Their goal is to cut through the hype and identify which use cases genuinely move the needle for productivity, business growth, and daily life.
💻 Computer Control & Automated Loops
- How it works: Andrew demonstrates telling Hermes to open VS Code, use the Claude Code plugin, and type a command to "loop" a task—essentially copying a competitor's entire business model. Hermes executes the command on the computer, controlling the desktop app directly.
- Speakers' opinions: Andrew is "super excited" about this, especially the looping capability. He sees it as a way to use unused Claude credits overnight, having Hermes work until a task is complete. Eric agrees but emphasizes the looping piece is more important than the initial command. He uses a "resolver" that observes his work across Claude Code, Codeex, and Hermes, then suggests what to loop and continue at the end of the day.
- Actionable takeaway: Use Hermes' computer control to automate repetitive coding or research tasks, and pair it with a resolver or loop system to ensure continuity and prioritization of work.
🔍 Competitor Research & Technical Breakdown
- How it works: Alex, another user, prompts Hermes to open a browser, go to a competitor's site (creatorbuddy.io), and perform a full technical breakdown—stack, features, payments, analytics, and what can be emulated. Hermes builds a complete report.
- Speakers' opinions: Andrew loves the delight on Alex's face and sees the potential to copy best features. Eric, however, calls this "useless by itself." He argues that one-off reports often go into the ether and are forgotten. The real value comes when you make it a repeatable cron job (daily, weekly, or monthly) that feeds into a channel or system where you actually take action.
- Actionable takeaway: Don't just run competitor research once. Set it as a recurring automation that delivers insights to a dedicated workspace (like a Slack channel or Notion database) so you're forced to act on the findings.
📅 Daily Routines & Task Triage
- How it works: A user wakes up, opens a Kanban board, and creates tasks like "script a new Hermes YouTube video" or "come up with three tweets." He then lets his Hermes agent take on these tasks.
- Speakers' opinions: Eric is highly critical, calling this "AI theater"—it looks productive but accomplishes little. He finds the Hermes Kanban board "useless" because it misses context and is hard to understand. Andrew agrees, preferring to put tasks in Notion instead. Eric's alternative is a "resolver" that organizes open threads from Claude Code, Codeex, and Hermes at the end of the day, giving a prioritized list of what to work on next.
- Actionable takeaway: Avoid flashy but shallow task management. Instead, use a resolver or daily priority list that consolidates your work across multiple AI tools, helping you focus on what truly moves the ball forward.
🗂️ Memory Wiki & Automated Diary
- How it works: A user prompts Hermes to build a "memory wiki"—a site with a list of all subjects discussed and daily logs of what was done. Hermes creates an automated diary that logs decisions and work over time.
- Speakers' opinions: Andrew uses Obsidian for daily logs and QMD (a GitHub memory tool) to make Hermes "less dumb." Eric uses QMD, GBrain, and GStack. He explains that QMD improves Hermes' ability to find information, while GBrain provides a stack of useful skills (CEO review, design review). Both agree that a closed-loop system—Obsidian for daily logs, Linear for project management, and QMD for memory—creates continuous improvement.
- Actionable takeaway: Upgrade Hermes' memory using tools like QMD or GBrain. Combine with a daily log (Obsidian) and a project management tool (Linear) to create a system that learns and improves over time.
📊 Project Management with Linear & Notion
- How it works: Eric uses Linear because of its simple API, allowing Hermes to create tasks that combine with Claude Code, Codeex, and the agent itself. Andrew uses Notion for his task board and has Hermes create tasks there when he finishes work in Claude.
- Speakers' opinions: Eric's setup worked well until his DGX Spark broke, but he plans to restore it. Andrew finds that passing tasks from Claude to Hermes via Notion works "flawlessly." He emphasizes the importance of having a way to get data from one AI tool to another so that action is taken.
- Actionable takeaway: Choose a project management tool with a good API (Linear) or one you already use (Notion). Set up a workflow where Hermes automatically creates tasks based on insights from other AI tools, ensuring nothing falls through the cracks.
💬 Communication Platforms: Slack, Telegram, Discord, & Zapier
- How it works: Andrew uses Telegram for private communication with Hermes. Brian Castle uses Discord for its markdown support and multi-channel capabilities. Andrew also uses Zapier MCP to give Hermes access to his email, automating follow-ups after interviews.
- Speakers' opinions: Eric finds Slack works fine for his team, even if threading isn't perfect. He argues that switching to Discord would be hard because his team is already on Slack. Andrew's Hermes advised him that Slack is a "9 out of 10" for multiplayer mode, while Discord is better for single-player. Eric also notes that getting a team to accept an AI agent on Slack is already difficult—switching platforms would make it harder.
- Actionable takeaway: Stick with the communication platform your team already uses. For single-player, Discord may be better; for team collaboration, Slack is the safer bet. Use Zapier MCP to give Hermes access to email for automated follow-ups and reminders.
🎥 Video Creation with Hyperframes & Overlap
- How it works: Nate Herk demonstrates Hermes creating a video using Hyperframes. Hermes runs terminal commands, uses vision to analyze the video, and iterates. The first pass is poor, but after asking to use Hyperframes, the second pass is much better.
- Speakers' opinions: Andrew thinks this is cool but notes the first pass wasn't great. Eric is more pragmatic: he believes future models (like a hypothetical "Fable 5") will make this effortless, but for now, he'd rather pay a human editor. He uses Overlap for content clipping, which finds mid-form clips that perform well on X and LinkedIn. Andrew adds that he hates text-only videos and would rather read at his own pace.
- Actionable takeaway: Video creation with Hermes is promising but not yet production-ready. For now, use tools like Overlap for automated clipping and save full video generation for when models improve. Focus on workflows that draft and iterate rather than expecting a finished product.
📰 Daily Briefings & Comment Monitoring
- How it works: A user runs cron jobs for daily AI news briefings, YouTube comment monitoring, morning business summaries, and follow-up reminders. Hermes responds to YouTube comments using the transcript and knowledge about the creator.
- Speakers' opinions: Eric admits he has a daily brief that he "doesn't read at all." He finds it helpful only for finding relevant AI business news for his marketing school. Andrew agrees—his OpenClaw has many crons he no longer reads. However, Eric sees value in having an agent reply to comments or analyze them to surface important ones. Andrew suggests using an agent to create downloadable resources from interview transcripts to drive email list growth.
- Actionable takeaway: Be selective with daily briefings—only keep those that directly inform your content or business decisions. For comment monitoring, focus on analysis and prioritization rather than automated replies. Use agents to create lead magnets (PDFs) from your content to grow your email list.
💡 Skills & Workflows: Saving and Iterating
- How it works: A user saves a workflow as a skill called "competitor-watch" so Hermes can run the same research on any topic without retyping the prompt. Andrew and Eric discuss how they update skills after big projects.
- Speakers' opinions: Andrew updates his skills manually after a project, but he doesn't get prescriptive—he lets Hermes handle it. Eric struggles with being too prescriptive and prefers a screen monitor that watches his work and suggests what to "skillify" at the end of the day. He finds that one-on-one conversations often contain gold that should be turned into skills. Both agree that overbuilding without handing off to a team leads to "dead open claws with a bunch of cron jobs."
- Actionable takeaway: Save successful workflows as skills, but don't overthink it. Use a screen monitor or daily review to identify what should be skillified. Most importantly, figure out who on your team can take over and run these skills so you don't become a bottleneck.
Final Takeaway
The most impactful Hermes use cases are those that move from one-off novelty to repeatable, actionable systems. Computer control and looping allow for overnight automation. Competitor research becomes powerful when scheduled as a cron job that feeds into a decision-making channel. Daily routines should be replaced with resolvers that prioritize your work across tools. Memory wikis and project management integrations create a closed-loop system that improves over time. Communication platforms should match your team's existing habits, and video creation is best left to specialized tools for now. Finally, skills and workflows must be handed off to a team to avoid becoming dead ends. The key insight from Andrew and Eric is clear: play and experiment, but always ask, "What's the next step? Who can take this and run with it?"