---
title: 'Build a Parallel Agent Workflow in 15 Minutes ⚚ Hermes Automation (No Coding!)'
source: 'https://youtube.com/watch?v=1MaFErWfL24'
video_id: '1MaFErWfL24'
date: 2026-06-16
duration_sec: 944
---

# Build a Parallel Agent Workflow in 15 Minutes ⚚ Hermes Automation (No Coding!)

> Source: [Build a Parallel Agent Workflow in 15 Minutes ⚚ Hermes Automation (No Coding!)](https://youtube.com/watch?v=1MaFErWfL24)

## Summary

This tutorial demonstrates how to build a parallel AI agent workflow using Hermes, enabling multiple agents to research different companies simultaneously and compile a unified report. The key innovation is using a single agent profile for a role, rather than creating separate agents for each company, which allows the system to scale dynamically by simply adding tasks.

### Key Points

- **Parallel Agent Philosophy** [0:00] — One AI agent handles one task at a time, but by splitting a job into pieces and running one agent on each piece simultaneously, you achieve parallel processing. The same agent role (e.g., researcher) can be assigned to multiple different targets (companies) without creating new agent profiles.
- **Correct Agent Setup: Role vs Topic** [1:08] — A common mistake is creating four separate agent profiles for four companies. The correct approach is to create one agent profile for the role (e.g., 'researcher') and assign tasks for each company to that same profile. This scales: adding more competitors only requires adding more tasks, not new profiles.
- **Agent Profile Configuration** [2:05] — Create a single profile named 'researcher' with a simple description: 'Researches a company's pricing and features.' Use a fast, cheap model like DeepSeek-V4-Flash for fetching and summarizing; the expensive model is reserved for the synthesizer that compiles the final report.
- **Adding Kanban Tool via Terminal** [2:51] — To enable the agent to create tasks from a Telegram message, the Kanban tool must be added. This involves backing up the config file, running a command to add Kanban to the tool set, then restarting the gateway. A conflict between two gateways can be fixed by asking Hermes to fix it via chat.
- **Task Creation in Kanban Dashboard** [4:58] — Create four empty tasks (one per company) in Kanban, then a fifth task for the synthesizer. The synthesizer task must have all four research tasks as its parents, so it only launches after all research is complete. The synthesizer prompt is crucial: it provides goals and output format (cover page, executive summary, charts, comparison table) but not step-by-step instructions.
- **Funneling and Automated Execution** [7:14] — The synthesizer waits for all parent tasks to finish before launching. Researchers are assigned to their respective tasks with company-specific prompts. The system automatically nudges tasks to start, and the parallel execution is faster than sequential human work.
- **Final Report Quality** [9:23] — The generated PDF includes a cover page, executive summary with a verdict ('ClickUp wins on price-to-feature ratio'), charts in brand colors, a full breakdown of pricing tiers, top features ranking, and a positioning matrix. The agent didn't just collect data—it made analytical judgments.
- **Full Automation with Memory and Cron** [10:49] — To automate daily runs, store the competitor list and builder instructions in the agent's memory via Telegram. Set up a cron job that pings the user every morning to confirm, then automatically creates the research tasks and builder task. Adding a new competitor is as simple as sending one chat message to update memory.
- **Dynamic Workflow Demonstration** [14:34] — Adding 'Linear' to the competitor list via chat updates memory without triggering research. When the user says 'Go', the system dynamically creates five research tasks (one per company) and the synthesizer, all without manual reconfiguration. The workflow reshapes itself around a single word added to a chat.

### Conclusion

By setting up a single agent profile for the researcher role and using Kanban to orchestrate tasks, you can build a parallel agent workflow that scales dynamically. The entire process runs automatically every morning, delivering a polished competitive analysis report to your phone with zero manual intervention.

## Transcript

One AI agent, one task at a time. That's
the ceiling, and it's slow. So, we break
it with Hermes. Run a whole team of
agents in parallel. So, you take one
job, split it into pieces, and put one
agent on each, all working at the exact
same time. Same role, different target.
Five agents, five companies all at once.
And you're never capped at five. Need a
10th next week? Text the bot one line,
and it's in. No setup, no new agents.
Then one last agent reads everything
they found and turns it into a single
report. And the whole thing is
automated. It runs every morning on its
own before you're even up. This is
what's on my phone when I wake up, a
full professional report. As you can
see, five competitors built and
delivered overnight. I didn't touch it.
No code.
Let me show you how to build this from
scratch.
So, we've got Kanban open.
We're building four agents that research
for companies at the same time. Then one
that turns it all into a report. So,
first, let's click on profiles,
where you set up your agents.
So, here's how I did it the first time.
Four agents, Asana, ClickUp, Monday, and
Notion. Four companies, four agents. And
that's wrong because an agent is a role,
not a topic. A researcher is a
researcher whether it reads Asana or
Notion. Same job, you're just pointing
it at a different company.
It's like hiring four people because
you've got four reports to write. You
don't. You hire one researcher and hand
them all four. And one researcher
scales. 10 competitors next week? Same
agent, more tasks. If you have four
profiles, you'd build a new one every
single time, and it's a dead end. So,
this go. Let's delete them.
And of course, we keep default. And now,
the right way. One profile. Let's click
on create.
Let's name it
researcher.
For the description, let's keep it
simple.
Researches a company's pricing and
features. And that's the role stated
once. Doesn't matter which company.
That's the whole point. The model is
DeepSeek-V4-Flash,
not the pro one, because these agents
just fetch and summarize. Flash is fast,
cheap, plenty for reading a pricing
page. The smart model comes in later,
where it earns it.
So, DeepSeek-V4-Flash and create. So, we
have one researcher, and that's the
agent doing all four jobs.
So, one more thing before we built.
Right now, the default agent runs the
bot, but it can't create tasks on its
own from a Telegram message, and we need
that at the end. So, we add the tool
now. Let's open our terminal, and let's
copy this command. So, this backs up the
config file, a copy, in case we break
something editing it. Anything goes
wrong, we restore from this. So, let's
run it. No output means it worked.
Okay, great. Our next command. Let's
paste it. So, this one adds Kanban to
the tool set list. The line that gives
the agent the power to create tasks.
Let's run it.
Great. Now, let's check.
And here we go. We added Kanban. So, we
have Hermes-CLI
and Kanban, both there. Now, let's
restart the gateway.
This will reboot the agent, so it loads
that new Kanban tool. Let's run it.
Okay.
Gateway is shutting down.
So, now we can see that start has
failed. Let's fix that. First, let's
type anything to Telegram, and it's
working.
But, we still have start failed. Let's
click on it.
Okay. So, the easiest way to fix this is
to copy this.
I know it sounds very lazy, but let's go
to chat.
Let's paste this here and tell Hermes to
fix it.
And we'll wait.
Ah, here we go.
It's running right now. I think it's
fixed. It's usually because it's a
Telegram conflict, two gateways fighting
over the same bot.
So, by telling Hermes to fix it, it will
close all the gateways and keep defaults
running.
Ah, here we go. Everything is fixed. You
can see the root cause. The researcher
profile's gateway was holding the
Telegram bot token, blocking the default
profile's gateway from starting. For
this kind of problems, it's better to
tell Hermes to fix it itself.
So, [snorts] setup's done. Now, we pawn
this one agent at four companies and let
it run. So, let's head over to Kanban.
Our dashboard. So, first thing to do is
to create the tasks, one per company.
So, first one is ClickUp.
Let's create, and then we have Monday.
Then Asana.
And lastly, Notion.
Okay. Four tasks, empty for now, no
assignee, no description, nothing. They
just sit there.
So, now the one that matters, which is
task five, the agent that pulls
everything together.
So, let's add it.
The name is the synthesizer. Let's
assign it to default.
And for parents, all of these tasks are
its parents. So, let's select one for
now. Let's create and let's open it up
because this one needs real prompt.
First thing's first, let's add more
parents, which is Monday,
Asana, and Notion.
And for the description, let me copy my
prompt and let's explain it.
So, this is the prompt that matters and
I want you to notice how it's written.
I'm not telling it which buttons to
press. I'm giving it the goal and the
output. Read what the four researchers
found, build a report, here's what the
report should look like. The design,
that's my brand colors, dark background,
gold fonts. I'm adding it a style guide.
The content, a cover, a takeaway, line
up top so it makes a point instead of
just dumping data, charts, a comparison
table, top five features. And the last
line, the media one, that's what drops
the finished file straight onto my
phone. What I'm not doing is telling it
how to install anything or which library
to use. I give it the what, it figures
out the how. That's what makes an agent
reliable instead of brittle. Okay? So,
we are done here. Let's save and let's
go back to our dashboard. And watch,
even with default assigned, the
Synthesizer does not launch. It just
sits because it's waiting on all four
parents to finish first. That's the
funneling. One agent holding until
everything it needs is done. Now, let's
arm the researchers. Each one gets the
same agent, researcher, and the same
instructions pointed at a different
company. So, let's start with ClickUp.
So, assignee is researcher
and let's edit the description. Save.
Second one is Monday, so researcher.
Let's edit it.
So, for ClickUp here, after
researcher, let's hit enter. Now, it's
done.
Asana
same thing
Let's edit the description.
And Notion So, researcher
And let's paste our command, our prompt,
and click save. Now, they start on their
own in about a minute.
The bot checks every 60 seconds, but I
don't want to stare at it. So, let's
click on nudge. And here we go. They're
in progress. So, the first one is
already done. Let's wait for the other
three.
So, while these run, here's the thing
people miss about agents. They're not
faster because they type faster, they're
faster because they work at the same
time. One of me does these four
back-to-back. Four agents do them at
once.
And they're finishing.
So, all four done, and now task five
should wake up. Let's scroll here.
So, it's in ready. Let's click on nudge.
And now it's in progress, and let's wait
for it.
It will be reading all four reports and
building the PDF.
>> [snorts]
>> First run, it installs the rendering
library, so it takes a minute.
And then the agent will solve its own
setup.
So, here, nothing for me to do. We'll
wait for the Telegram message.
And there it is, straight to Telegram. I
didn't write this, didn't design a
single page for agent research. One
agent built it, and it landed on my
phone as a finished file.
So, let's open it.
And look at this. A cover page, title,
the four platforms, the quarter, the
date. This isn't a text dump. It's a
document. And right here, page two, the
executive summary.
Read that first line, ClickUp wins on
price-to-feature ratio.
That's not data, that's a verdict. The
agent didn't just collect numbers, it
took a position.
I come back to that because that's the
most important thing in the whole
report. And the chart, two of them,
entry tier and most popular tier, side
by side in my colors, the gold, the dark
background, it built those from the
numbers the researchers built. Then we
have the full breakdown, every tier,
every price for all four, ClickUp,
Asana, monday, Notion, what each plan
costs and what you actually get. And you
have here top five features across
platform ranked, AI agents, automations,
the view types, integrations, enterprise
security one to five. On the last page,
this is the one, a positioning matrix.
Each platform, its price range, its
wrist strength, what it's for, ClickUp
for cost-conscious teams, Asana for
regulated enterprises. This is strategy,
not a spreadsheet. So, for agents
research this, something you'd actually
send a client. But there is one more
thing I still had to do. Sit here and
start it. So, let's fix that. Let's make
it an automated workflow where you can
add competitors and set that to work
every day.
So, to do that, let's go first to
Telegram. So, the first thing to do is
to give the bot a memory, my competitors
once right here in the chat. So, let's
paste this command. Okay, memory my
competitor list, ClickUp, monday, Asana,
Notion. Remember this, I'll ask you to
run research on them later. Let's run
it.
And now it knows my list and I can
change it anytime just by talking to it.
You'll see. So, one more thing in its
memory, the builder's instructions. So,
when it runs, it knows exactly how to
assemble the report. This is the brain
behind that PDF. Let's paste this, and
let's run it.
So, now we have the list and the builder
both in memory. Now, the schedule. I
want it to ping me every morning and ask
if I want to run. And here's the catch,
a scheduled job can't ask me and then
sit there waiting for an answer. It runs
alone, no chat. So, it pings me and I
reply separately. Ping, then reply. So,
let's paste our cron command, and let's
run it. And that's the whole engine in
one instruction. The ping only asks, but
go builds everything. A research task
per company and the builder on top
waiting on all of them. And I made it
confirm the count back to me, so I
always know the builder's there. So, I'm
not waiting until 9:00 a.m. tomorrow to
show you. I'll tell it to run the job
right now. So, let's say, "Run the daily
search ping job now." And let's run it.
So, let's wait for it. It's running, and
there's the ping, "Morning. Here's your
list ClickUp Monday Asana Notion.
Edit or say go." And it pulled that from
memory. I didn't paste it. So, let's
say, "Go." And then enter.
And it confirms. So, it's creating the
four tasks, as you can see here. And
then the synthetizer. So, let's go to
the board. Here you go, four researchers
and one synthetizer. Great.
So, now let's wait for our PDF, and
after that we run a new competitor.
Great. So, everything is working. Now,
the part that makes this useful, the
dynamic workflow, because our list isn't
locked. When a new competitor shows up,
we just tell it to add it. For example,
let's add Linear to the list.
So, add Linear to my competitor list
just update the memory list. Do not
create or run anything yet. Let's run
it.
And it just updates the memory. Nothing
runs. Adding a name and running the
research are two different things. It
knows the difference. Lists, we have
five competitors now.
And the board hasn't moved.
Now, let's say go.
And hit enter.
So, five research tasks now and the
builder on top of all five. So, I have
six.
So, let's see the board. So, we have
five research tasks and the synthesizer,
the sixth one.
So, five researchers, one for each
company including Linear, and the
builder waiting on all five. Great.
So, I added one word to a chat message
and the whole machine reshaped around
it.
That's an agent, not a script. I didn't
rewire anything. I told it.
So, here is where we landed. Every
morning at 9:00, it pings me with my
list. I reply, edit it, or just say go.
And it builds the whole thing. Finds
out, researches everything, sends me the
finished report.
I set it up once. Now, it runs itself.
And here's the final report. Same
engine, but now with Linear in it. Let's
click on it.
And right on the cover, we have ClickUp,
Asana, Monday, Notion, and Linear.
If you scroll down,
we can see Linear here as part of the
chart. So, it didn't fill a slot. It
researched Linear and took a position on
it. That's the whole thing. Five
competitors, one word added to a chat,
four reports on my phone, and tomorrow
at 9:00 it asks me again.
And that's parallel agents in Hermes.
We're all many targets running on their
own every morning. If you want to run
Hermes 24/7, there is a VPS link in the
description. Use the code nocodehack for
a discount. Like the video, subscribe
for more Hermes builds, and I'll see you
in the next one.