Video summary

How To Set Up Hermes Agent To Run Your Day

Main summary

Key takeaways

Technology

Tech/Product Concepts Covered

  • Hermes agent is more than a chatbot: it supports memory, tools, scheduled jobs (cron-style automation), skills, sub-agents, file handling, and browser/computer access to execute work across your setup—not just respond to prompts.
  • Core setup philosophy: the biggest mistake is trying to feed Hermes random one-off tasks immediately. Instead, Hermes should first learn your actual recurring responsibilities and bottlenecks so it can generate relevant workflows.
  • Delegation model / safety boundaries: create a delegation map that categorizes tasks by risk level and required approval:
    1. Fully automated (no approval)
    2. Draft/research only (approval before execution)
    3. Reminder/context only
    4. Too risky—manual

Guide/Tutorial Structure (Step-by-Step System)

Step 1 — Hermes “Life Audit” Interview Prompt

  • Hermes interviews the user one question at a time about:
    • work
    • recurring tasks
    • decisions
    • messages
    • research/content/admin
    • personal context
  • After the user answers, Hermes outputs:
    • top 10 tasks Hermes can help with
    • three easiest workflows to set up today
    • three highest-leverage workflows for the week
    • required tools/files/permissions/context for each
    • a daily check-in prompt to schedule
  • Emphasis: avoid generic productivity suggestions; workflows must be grounded in what the user actually reports (e.g., founder → investor updates; dev → code review/triage/release notes; creator → titles/thumbnails/video ideas).

Step 2 — Build a “Delegation Map”

  • Hermes takes everything from the interview and organizes potential tasks into the four delegation/safety sections above.
  • For each task category, it explains:
    • why it belongs
    • how to do it well
    • what the first version should look like
    • safety boundaries
    • how to know the workflow is working

Step 3 — Pick Only 3 Workflows (Don’t Automate Everything at Once)

Choose exactly:

  • one easy workflow
  • one useful recurring workflow
  • one high-leverage workflow

Rationale: keep setup small so it’s actually used and refined.

Example workflows suggested in the video (from the author’s scenario):

  • “X reaction packet”: surfaces X content to react to (addressing the user missing posting/reacting).
  • “Pre-PM call client brief” from Fireflies transcripts: summarizes call trajectory and flags issues to check.
  • “Mini client risk digest” for 3 clients: validates a monitoring/reporting workflow before scaling to all clients.

Step 4 — Morning Check-In Cron (Daily Loop)

  • Schedule for 9:00 a.m. every day:
    • ask the user’s #1 priority for today
    • summarize the goal in one sentence
    • suggest three ways Hermes can help (based on the delegation map)
    • ask whether to start any tasks
    • if answers imply durable preference/recurring responsibility, suggest a memory update (only if it will matter later)
  • Strong constraint: keep it concise to avoid overwhelming the user.

Step 5 — Create a Short “Operating Profile” (Context for Better Outputs)

  • Provide Hermes a one-page operating profile (or generate it from the interview).
  • Include only what matters now and next month:
    • what you do
    • tools
    • recurring responsibilities
    • frequent decisions
    • disliked tasks
    • explicit preferences/constraints
  • Warning: avoid a “memory junk jar”—don’t store every detail; store only high-signal facts.

Step 6 — Convert Repetition into “Skills” (SOPs)

  • If the same workflow/research process/report repeats, turn it into a skill (reusable SOP).
  • Example skill process described for YouTube:
    • check posted videos
    • check what’s already filmed
    • scan competitors for outliers
    • pull transcripts
    • use VidIQ
    • improve titles
  • Claimed result: Hermes becomes more effective over time (e.g., better by day 30) by refining and reusing these SOPs.

Other Practical Product Notes

  • The video mentions existing infrastructure for the demo use case:
    • a client experience dashboard in MaxHQ (with GitHub linkage)
    • snippets in Notion
    • used as a data source for workflow auditing/monitoring
  • After setting up delegation + daily check-in, improvement depends on:
    • Hermes having context
    • correct tools/permissions
    • user preferences

Main Speakers / Sources

  • Primary speaker: the tutorial author (narrator) guiding the Hermes setup.
  • Referenced sources/tools (not speakers):
    • Hermes agent capabilities
    • Fireflies (call transcripts)
    • MaxHQ (dashboard)
    • VidIQ (YouTube analytics)
    • GitHub
    • Notion
    • Cron scheduling

Original video