10 Hermes Agent Hacks That Turned My Chat Agent Into a 24/7 System

Overview

Most people use Hermes Agent like a chat app: open it, type a prompt, get a response, close it. That leaves roughly 90% of what Hermes can do on the table.

These ten setups turn Hermes from a chat window into a 24/7 system that works while you sleep, reacts when your workflow changes, and gets sharper with every run. They saved the author 15+ hours every week, and they work for any workflow you run repeatedly — content, software development, business operations, client management, research, or sales. If you do it more than once, Hermes can run it.

The examples below come from content and social media automation because that's what the author runs daily, but the mechanics are domain-agnostic. A cron job that scans X for trending posts uses the same mechanic as one that checks GitHub for open PRs or monitors a CRM for new leads.

If you only have time for three, start with: Cron Jobs (#3), structured /goal (#4), and Skills (#8). These three alone change how Hermes feels overnight.

Setup time and time saved

1. Mission Control

The first and biggest setup: build a dashboard where everything is visible. When Hermes is doing real work, you don't want that work buried inside a chat thread. You want to see what's running, what's waiting on you, what's blocked, what needs approval, and what changed since yesterday.

Ask Hermes to build it:

Build me a mission control dashboard. Start with:
- A kanban board showing all active agent tasks
- A content pipeline where I can add ideas and track progress
- A memory wiki showing everything we've worked on
- A performance section showing my X and content metrics

Hermes also ships with a built-in dashboard out of the box:

hermes dashboard

It opens at localhost:9119 with skills, models, cron jobs, profiles, and the kanban board. Start there, then customize when you need more.

Hermes also launched a native Desktop app for macOS, Windows, and Linux — side-by-side preview, file browser, and integrated voice, sharing the same data directory as the CLI and Telegram. Work from Desktop at your machine, switch to Telegram on the go. One agent, every surface.

Fastest path to a working agent on any surface:

hermes setup --portal

One OAuth covers the model, web search, image generation, TTS, and cloud browser — no separate API keys needed. Once a Kanban board lives directly in the dashboard, Hermes stops being something you message and becomes part of your operating layer.

2. Event Triggers

Think about where you already work: Notion, Linear, Google Sheets, Slack. You move a task, update a status, add an idea — and right now, nothing happens after that. You have to remember to tell Hermes about it later. The fix: make Hermes watch for changes and react automatically.

Example workflow: when you move a video idea to your "To Film" list in Notion, Hermes detects the change and sends a filming brief to Telegram within minutes — including whether to film now/later/kill, the strongest title angle, a 30-second hook, the proof assets you'll need, and a pre-filming checklist. You didn't prompt Hermes; you moved a card.

Option A — cron job watches for changes (simplest). Schedule a job every 10 minutes:

check my Notion board [board URL].
if any card moved to "To Film" in the last 10 minutes,
research the topic, write a filming brief,
send it to Telegram.

Option B — webhook trigger (instant). Use Notion automations, Make, or Zapier to send a webhook to Hermes when a card moves. The response is instant instead of polling every 10 minutes. The principle: when your workflow changes state, Hermes should know what to do next.

3. Cron Jobs

Event triggers react to changes; cron jobs react to time. Every morning you get useful information before you ask for it — that shift makes Hermes feel like an employee who starts work before you wake up.

Every morning at 8am:
send me one AI story worth reacting to on X.

Every 3 hours:
scan X for fresh posts in my niche I should quote tweet.

Every day at 9pm:
check if competitors posted any outlier content today.

Every Monday at 9am:
audit my content board. flag ideas stuck for more than 7 days.

Every Friday at 6pm:
summarize what content shipped this week,
what performed, what didn't, and why.

Setting these up is plain English — no crontab syntax. Just tell the agent what you want and when. Useful information arrives before you even think to ask.

4. /goal With Structure

A normal prompt asks Hermes for one response. /goal gives Hermes an objective to work toward across multiple turns until it's done. Most people use /goal like a prompt — vague in, vague out. The difference between a useless /goal and one that ships real work is structure.

/goal [OUTCOME]
using [SOURCES]
with constraints: [CONSTRAINTS]
deliverable: [DELIVERABLE]

Each part does a job:

• Outcome tells Hermes when the goal is achieved
• Sources tells it where to look
• Constraints tell it what to avoid
• Deliverable tells it what "done" looks like

The interview hack. If you don't know how to structure your goal, make Hermes do it for you:

I want to use /goal but I don't want a vague goal.
Interview me with only the questions you need.
Then turn my answers into the strongest possible
/goal command. Include the exact outcome, context,
sources, constraints, deliverable,
and when you should stop.

Hermes asks 5-8 questions, then writes its own /goal command from your answers — sharper than anything you'd write from scratch.

5. Sub-Agents as a Research Team

One agent gives you an answer; sub-agents give you a team. For any research task worth doing, split it across multiple sub-agents running in parallel, each with a different source, and merge the results into one recommendation.

/goal research the best content angle for this week.
spawn 3 sub-agents:

1. scan X for trending posts in AI agents niche,
   pull engagement numbers and hooks that worked

2. analyze my last 30 days of posts,
   find patterns in what performed vs what didn't

3. check competitor accounts,
   flag any outlier content from the last 7 days

combine all three into one recommendation
with the strongest angle, a draft hook,
and proof assets I'll need.

Each sub-agent gets its own context window; only the final summary returns to the main session, so your main context stays light. Best use cases: research across multiple sources, competitive analysis (one sub-agent per competitor), content creation (research/draft/edit), and code reviews (logic/security/performance).

6. Telegram Topics as Workspaces

Telegram topics turn one chat into separate workspaces, each with a different context and job:

• YouTube — content planning, scripts, filming briefs
• React — trending posts on X worth reacting to
• Coding — technical work, debugging, PRs
• Research — deep dives, competitor analysis
• General — smaller tasks, random questions

When everything runs in one chat, context bleeds — a coding question gets mixed with a content brief. Topics fix that; Hermes knows what you're talking about based on which topic you're in.

To set up: create a group with your Hermes bot, enable Topics in group settings, create a topic per workspace, then message Hermes in each topic separately. Each topic can get its own cron job:

React topic cron, every 3 hours:
scan X for posts in AI agents niche
with 500+ likes in the last 3 hours.
if any are worth reacting to, draft a quote tweet
and send it here for approval.

Research stays in Research, content stays in Content — no cross-contamination.

7. Kanban for Task Management

Once Hermes works on more than one thing, you need a board, or tasks disappear into chat. Hermes has a built-in Kanban board with durable SQLite storage, shared across all profiles.

hermes kanban list

Drop tasks into triage and the dispatcher auto-assigns them to workers every 60 seconds. Statuses flow: Triage → To-Do → Ready → Running → Blocked → Done. You see what's ready, running, and done; which agent owns which task; and what's blocked and why. Crashed tasks get auto-reclaimed (zombie detection) and heartbeats track worker health.

Every /goal you set also becomes a Kanban card automatically:

/goal research competitors → kanban card
/goal draft weekly report → kanban card
/goal triage inbox → kanban card

Drop five tasks at breakfast; by lunch, half are done — and you didn't manage any of them.

8. Skills as SOPs

A skill is a standard operating procedure for Hermes: encode a process once and the agent uses it forever. Hermes already creates skills on its own after every task — it reviews what worked, saves the workflow as a markdown file in ~/.hermes/skills/, and reuses it next time. Writing skills intentionally for your key workflows is where the leverage multiplies.

Save this as a skill called "content-post":

# Content Post Workflow

1. Check trending topics in AI agents niche via X search
2. Cross-reference with my last 14 days of posts (avoid repeats)
3. Pick the strongest angle based on engagement patterns
4. Write a draft in my voice
5. Score the draft:
   - Hook: does it stop the scroll? (1-10)
   - Bookmark fuel: would someone save this? (1-10)
   - Proof: is every claim backed by a number? (1-10)
6. If any score below 7, rewrite that section
7. Send final draft to Telegram for approval

Now whenever you say "use content-post for today's draft," Hermes runs the entire SOP without you explaining it again. Any workflow you explain twice should become a skill. Skills are transparent — they live as markdown files you can read, edit, or delete. No black box.

hermes skills

Hermes ships with 60+ built-in tools across terminal, web, browser, vision, image generation, TTS, and code execution. Skills layer on top of those tools to create full workflows.

9. Webhooks and Event-Based Agents

Cron jobs run because the clock changed; webhooks run because the world changed. Examples of event-based triggers:

• A new lead comes in → Hermes researches the company immediately
• A GitHub PR opens → Hermes summarizes the changes and flags risks
• A competitor posts content → Hermes checks if it's worth reacting to
• A meeting transcript drops → Hermes extracts action items and adds tasks to your board
• A keyword starts trending → Hermes drafts a content angle

Hermes receives webhooks through its gateway. Configure the webhook URL in your automation tool (Make, Zapier, n8n) and point it at your Hermes gateway endpoint.

n8n workflow:
1. RSS trigger watches competitor blog (every 30 min)
2. if new post detected → send webhook to Hermes

Hermes /goal on webhook receive:
/goal a competitor just published: [title] [url].
read the full article via web search.
summarize the key points in 3 lines.
assess: should I react to this on X?
if yes, draft a reaction post in my voice.
send everything to Telegram for approval.

The principle: cron jobs handle time, webhooks handle events. Together they cover every scenario where Hermes should wake up without you touching it.

10. Separate Agents by Job

You don't want one agent doing every job with the same model, tools, memory, and permissions. Hermes profiles let you create separate agents for separate roles, each with its own soul.md (personality and rules), memory, skills, model, MCP connections, and permissions.

hermes profile create content-lead
→ soul.md: you produce content. match my voice.
   use trending data. avoid repeated angles.
→ model: strong writing model
→ tools: X search, web search, analytics

hermes profile create researcher
→ soul.md: you find information. deep research only.
   no opinions. facts and numbers.
→ model: cheaper, high-volume model
→ tools: web search, firecrawl, browser-use

hermes profile create ops
→ soul.md: you handle admin. calendar, email triage,
   reminders. ask for approval before sending anything.
→ tools: email, calendar, notion

hermes profile create code-reviewer
→ soul.md: you review PRs. flag security issues,
   logic errors, performance problems.
→ model: deep-reasoning model
→ tools: github, terminal

Some agents need the smartest model you can afford; some just check a page every hour. Some should have write access; some never should. Each profile runs its first /goal, learns from the result, and saves the workflow as a skill — the second run is faster, the fifth is automatic.

Share any profile with one command:

cd ~/.hermes/profiles/researcher
git init && git add . && git commit -m "initial"
git push origin main

Anyone can install it with hermes profile install github.com/you/researcher. They fill in their own API keys; their memories and sessions stay separate.

How They Chain Together

These setups compound when stacked. One chain running in the author's system:

8:00 AM — cron job (#3) fires.

the content-lead profile (#10) wakes up
and starts a structured /goal (#4):
"find the 3 strongest content angles for today
using X trending data and my last 14 days of posts."

it spawns 3 sub-agents (#5):
→ sub-agent 1 scans X for trending posts
→ sub-agent 2 pulls my recent post performance
→ sub-agent 3 checks competitor accounts

all three become kanban cards (#7).
dispatcher tracks them in parallel.

sub-agents finish. content-lead runs
the content-post skill (#8) to draft 2 posts.

drafts land in my Content topic
in Telegram (#6) for approval.

I tap approve on one. reject the other.

10 minutes later a competitor publishes
a reaction. a webhook (#9) fires.
Hermes drafts a follow-up angle
and sends it to my React topic (#6).

I see everything on mission control (#1).

One morning. Seven hacks fired. Two posts ready. Zero manual research. That is the system.

The Real Insight

If Hermes still feels like another chat app, look at the system around it. Give it a mission control so you can see what's happening. Set up event triggers so it reacts when your workflow changes. Add cron jobs so useful information arrives before you ask. Use /goal with structure instead of vague prompts. Split research across sub-agents. Separate workspaces with Telegram topics. Track tasks on the Kanban board. Turn repeatable processes into skills. Connect outside events via webhooks. Stop making one agent do every job.

Ten setups, each saving hours per week. Stack all ten and Hermes runs your operations while you focus on the work that moves the needle. The agent is ready, the stack is ready — wire the system and let it work.

Flow contributed by YanXbt. For the official reference, see the Hermes Agent documentation.