Video summary
Hermes Agent Use Cases That Are Actually Insane
Main summary
Key takeaways
Tech/Product Summary (Hermes Agent)
Hermes positioning & comparison
- Hermes is described as more capable than “Open Claw.”
- The comparison emphasizes that Hermes includes extra built-in features that enable broader automation and optimization.
New UI: Desktop app (wrapper around agent setup)
The desktop app is presented as an upgrade over the terminal UI (TUI):
- Provides a more interactive experience aimed at easier:
- Monitoring of agents
- Configuration
- Lets users configure:
- Model
- Tools
- Enabled skills
- Via a settings pane
- Supports memory cap / profile budgets using a character-count-based limit (described as similar to the previous video).
- When switching from TUI to the desktop app, it “picks up those same configs” (existing TUI configuration is carried forward).
Multiple profiles (key capability difference)
- A profile = an individual Hermes agent instance with:
- Separate memory
- Separate skills
- Used to isolate different “personas”
- The desktop app can run several profiles in parallel, delegating tasks concurrently.
- On the terminal, switching/managing profiles allegedly requires many commands; the desktop UI simplifies this.
Core Automation Features & “Flags”
1) Wake agent feature (cost-control, event-driven LLM use)
- Implemented as a flag checked during cron execution.
- If wake agent = false:
- the LLM call is skipped for that event.
- Benefit:
- prevents wasting LLM tokens/cost on cron triggers that don’t require AI attention.
- Contrast:
- Open Claw is reportedly run on every cron job regardless of necessity.
Example workflows
- AWS cost monitoring + Gemini API monitoring
- If costs rise above normal, report to a channel.
- Stripe MCP (and other tools) for threshold-based actions
- Take actions when thresholds are exceeded.
- Google Play Store review management
- Uses the Google Play API with cron.
- Goes beyond reacting to one-star reviews:
- Reports in Slack
- Identifies common issues
- Helps teams interpret user sentiment and trends.
2) No agent flag (non-AI cron jobs inside Hermes)
- A separate mode where scheduled tasks run without invoking the AI model.
- These tasks still run inside the Hermes ecosystem, with easier setup because Hermes has richer workflow context.
- Upside:
- described as “practically free” (minimal/no agent/LLM cost).
Example
- Monitor site health via TLS metrics plus Stripe app health, then post updates to Slack.
- Hermes can also generate cron automation from prompts, using the required skills.
Organizational “Second Brain” Use Case (Slack-Based Team Hub)
- Connect Hermes to Slack so team members can interact with it directly.
- Focus is on:
- Evolving memory
- Evolving skills
- Intended effect:
- observe day-to-day work and convert it into reusable workflows.
How it’s described to work
- Accumulates context about the organization, tasks, and progress.
- Creates a company skill that encapsulates organizational operations knowledge.
- Enables a shared central hub:
- team members ask questions
- check status
- Project management example:
- track goals/progress
- assign tasks
- update teammates based on goals
Long-running task durability
- Hermes is purported to handle ongoing interactions better via memory editing.
- Open Claw is criticized for “soul file bloating” over time, requiring resets to remain effective.
Built-in Skills / Actions (Email, PRDs/Competitor Analysis, Social Media)
Gmail lead response automation
- Hermes includes built-in skills for Google Workspace access (via scripts).
- Setup steps described:
- Create a Google Cloud project
- Obtain credentials
- Add a callback token to connect Hermes to Gmail
- Automation mechanism:
- Hermes creates a webhook to monitor incoming emails
- It analyzes emails for “potential leads”
- It auto-generates replies using stored company context
- If needed, it can schedule meetings based on calendar availability
PRD as a skill + competitor-driven updates
- Emphasis: creating a PRD as a skill (not just a document).
- Claimed benefits:
- Skills are loaded only when needed
- The PRD remains in the model’s active context window
- Reduces risk of “losing track” of requirements
Competitor workflow
- Hermes can analyze competitors using the PRD skill context.
- A weekly cron job:
- studies competitors
- updates the PRD with feature ideas
- maintains a separate competitor-progress document
- These files become part of the agent context so recommendations can directly feed product building.
- Motivation:
- reduce manual monitoring (e.g., “doomscrolling” for updates)
Social media repurposing with a human review step
- Mentions a skill called XURL to automate posting to X.
- Workflow:
- Provide a video script
- Hermes generates X and LinkedIn posts into a specified folder
- A human review/edit step ensures accuracy and tailoring
- A slash command triggers posting via the skill
- X posts must respect character limits.
Reviews / Guides / Tutorial Emphasis
- Desktop app vs TUI is framed as a guide/UX improvement:
- “more interactive”
- easier monitoring/config
- Installation/config is presented as a how-to:
- download installer for the OS
- configure model/tools/skills
- Wake agent and no-agent flags are explained as practical tutorial patterns for cost-effective automation.
- Multiple step-by-step implementations are described, including:
- Slack second-brain setup conceptually (connect bot to workspace)
- Gmail connection (Google Cloud credentials + callback token)
- Webhook-based lead email reply flow
- Weekly competitor cron job + PRD update mechanism
- Social repurposing flow:
- draft to folder → human approval → publish
Main Speakers / Sources
- Primary source/speaker: “we” / channel host (likely the YouTube creator for AI Labs Pro / Hermès-focused content; no individual name given in the subtitles)
- Mentioned external entities:
- Brilliant (sponsor)
- Claude code (mentioned in context of using Claude with Hermes)