You can build a fully autonomous Spotify DJ agent in about seven minutes with Hermes, and you do not need to write a single line of code. You tell it the mood you want in any chat window, it reads the vibe, searches Spotify, and queues the songs, and a couple of cron jobs keep your playlists fresh while you sleep. Here is exactly how to build it.
I have been running mine for 714 consecutive cron runs without a single failure. It backfills my Bruce Springsteen and Bob Dylan playlists every day at 5pm, and the only time I think about it is when a “Added 2 tracks” message lands in my Telegram. This guide walks you through the same DJ profile setup, the three custom skills that turn a basic music bot into a real tastemaker, and the cron jobs that make it run forever.
Core Concepts
- A Hermes profile is an isolated agent with its own model, personality, and connections, so your DJ stays separate from everything else you run.
- The base Spotify skill gives you 7 tools: play, skip, volume, search, queue, playlists, and library access.
- Custom skills are where the magic lives. Vibe mining, melodic house curation, and a sampler platter turn search into real music discovery.
- Cron jobs let the agent run on autopilot, backfilling playlists daily with zero input from you.
- It works in any chat surface you already use: Telegram, Discord, SMS, or the terminal.
Who This Applies To
This is for anyone comfortable in the terminal who has not used Hermes yet. If you can run a command and edit a config file, you can build this. You do not need to be a developer, and you do not need to touch the Spotify app once it is running.
What Is a Hermes DJ Agent?
A Hermes DJ agent is a dedicated AI agent that controls your Spotify for you from any chat window. You type something like “play something for cooking dinner,” and it picks the right vibe, queues the tracks, and starts playback, all without you opening Spotify. Because Hermes runs as a persistent agent, you can also schedule it with cron jobs so it manages your playlists on a daily loop.
The reason it feels different from a normal music bot is that Hermes treats your DJ as its own walled-off profile. That profile carries its own model choice, its own personality, its own Spotify authentication, and its own custom skills. Nothing leaks between agents, and you can spin up a totally different agent for email or research without any of them stepping on each other. This is the same agent model behind the broader shift toward open-source AI agents replacing traditional software.
🎧 From my setup: The thing that sold me was the autopilot. I wanted a Bruce Springsteen catalog seeded back into rotation. Instead of doing it manually, I let a cron job scan my recently played tracks and backfill the playlist every day at 5pm. 714 runs later it has never missed, and I have never opened the playlist editor again.
What You Need Before You Start
The requirements are short, and you are about five minutes from a working DJ.
- A Mac or Linux machine with Hermes installed. Confirm it is ready by running
hermes --versionand looking for a clean version number. - A Spotify Premium account. This is the one hard requirement for playback. Spotify Free works for search and playlist building, but it will not let the agent actually play audio. You can grab a Spotify Premium account here.
That is the whole list. No code, no plugins, no extra services.
How Do You Create the DJ Profile?
You create the DJ profile with a single command, then edit two lines in its config. Hermes uses profiles to keep agents isolated, so the first step is to spin up a fresh one for your DJ:
hermes profile create djThat gives you a clean profile at ~/.hermes/profiles/dj/. Open its config.yaml and change two things:
model: deepseek-v4-flash # fast and cheap, perfect for short music commands
personality: kawaii # gives the DJ a playful voiceThe model choice matters more than it looks. Music commands are short and frequent, so you want something fast and inexpensive. deepseek-v4-flash is ideal here. If you want to see how far that cost efficiency scales, I broke down how to run a multi-job AI automation pipeline for pennies a day using the same model.
The personality is pure flavor. The kawaii personality makes the DJ feel playful and fun, but any personality works. Try surfer, noir, or build your own. It only changes the voice of the responses, not the music logic.
Now connect Spotify with one command:
hermes auth spotifyA browser window pops up. Click Agree, and you are done. Switch back to the terminal and you will see a green “Spotify connected.” This is a one-time step, and the auth lives inside the dj profile, so it stays put even as you add skills and cron jobs later.
The Secret Sauce: Three Custom Skills That Make It a Real DJ
The base Spotify skill handles the mechanics: play, search, queue, playlists, and the rest of the 7 tools. That is enough to control music, but it is not enough to discover music. The real magic lives in the skills directory at ~/.hermes/profiles/dj/skills/, where you layer custom skills on top of the base. Here are the three I run.
Vibe Mining: The Core Discovery Engine
Vibe mining is the skill that finds music that sounds like a reference track. You hand it a seed, and instead of returning one obvious result, it searches in concentric rings to map the whole ecosystem around that sound. It runs up to four parallel searches per round across distinct patterns:
- Big-name artists in the same lane
- Curated playlists that match the mood
- Underground artists most engines skip
- Deep cuts and lesser-known tracks from the catalog
Inside the SKILL.md, the multi-pass search section pairs with a “vibe pocket” table that maps moods to subgenres like melodic house, deep melodic, organic house, and melodic techno. Give it a seed track and it returns the whole ecosystem around that sound, not just the one song you already knew.
Melodic House Curation: Niche Discovery
Where vibe mining is broad, this skill goes deep on a single genre: melodic progressive house. Think Albert Klein, Ben Böhmer, and Lane 8 territory. The difference is that the curation logic is baked right into the skill. It carries curated playlist IDs from labels like Anjunadeep, HMWL, and Klangspot, and it prioritizes underground artists with fewer than 50,000 monthly listeners, names like Melosense, Klur, and Archaellum.
The payoff is that you get a personal tastemaker for one genre. Instead of recycling the same top tracks, it surfaces the artists you would only find by digging through label playlists for hours.
Sampler Platter: Compare Playlists Side by Side
Sampler platter answers a simple question: two playlists, one vibe, which one wins? It queues tracks from multiple playlists so you can compare them side by side, or it curates a themed journey that pulls across your whole library. The skill ships with two modes:
- Variant A, Compare and Choose: interleaves tracks from each playlist so you can A/B them in real time.
- Variant B, Theme Journey: builds one flowing sequence across multiple sources, with emoji section headers marking the shifts.
There is also a “Phase 2, beyond the library” section that pushes past your saved music into fresh discovery once you have decided what you like.
A Real Use Case: Building a 3-Hour Airport Playlist
Here is the exact thing I built this for. Before a flight, I used to spend twenty minutes assembling a long travel playlist by hand, hunting for tracks that flowed together so I had something to play door to door without skipping. Now I just ask the DJ.
I tell it to build a roughly three-hour playlist mixing artists like Nora En Pure and Ben Böhmer, and it does the heavy lifting. Vibe mining maps the ecosystem around those two as the seeds, pulling in the same melodic and organic house lane, and the sampler platter’s Theme Journey mode sequences it all into one flowing set instead of a random shuffle. A few seconds later I have a cohesive three-hour playlist that actually holds a mood the whole way through.
The last step is the only manual one, and it lives in the Spotify app: open the new playlist, flip on Download, and it saves offline to my phone. Airplane mode for the entire trip, no buffering, no signal needed at 35,000 feet. One sentence to the agent, one toggle in the app, and the full flight is covered.
✈️ Why this beats doing it by hand: A manual three-hour playlist means scrolling, second-guessing the order, and usually padding it with tracks you already know. Seeding the agent with two artists you trust lets it surface the deep cuts and adjacent artists you would never have queued yourself, so the playlist stays fresh for the whole trip instead of looping your usual rotation.
Set It and Forget It: The Cron Jobs That Run Forever
The real flex is that none of this needs you once it is built. Hermes cron jobs let the DJ profile run on a schedule, scanning your recently played tracks and backfilling playlists automatically. The job definitions live in ~/.hermes/profiles/dj/cron/jobs.json.
Here is my Springsteen job, which runs daily at 5pm:
{
"name": "springsteen-backfill",
"schedule": "0 17 * * *",
"state": "scheduled",
"completed": 714
}The 0 17 * * * is standard cron schedule syntax for “every day at 5pm.” A second job syncs Bob Dylan tracks at the same time and has logged 57 runs so far. Both run inside the dj profile, so they inherit the same Spotify auth and skills you already set up.
The headline number: 714 runs without a single failure. Every afternoon, a quiet “Added N tracks” message streams into Telegram, and the playlists stay fresh on their own. Build it once, and it runs forever. If you want another practical example of this pattern, here is how to set up daily news brief cron jobs with Hermes.
Manual Spotify vs. a Hermes DJ Agent
Here is the quick side-by-side on why this is worth the seven minutes.
| Task | Manual Spotify | Hermes DJ Agent |
|---|---|---|
| Start a vibe | Open app, search, build a queue by hand | Type “play something for cooking” in any chat |
| Discover new music | Rely on the same algorithmic radio | Vibe mining searches 4 rings, including underground |
| Niche genre digging | Hunt through label playlists for hours | Curated skill surfaces sub-50K-listener artists |
| Keep playlists fresh | Remember to update them yourself | Cron jobs backfill daily, hands-off |
| Where it runs | The Spotify app only | Telegram, Discord, SMS, or terminal |
Frequently Asked Questions
Do I need to know how to code to build this?
No. The whole setup is two commands, two lines in a config file, and dropping skill files into a folder. Hermes does all of this for you. If you can use the terminal, you can build this in about five to seven minutes.
Does it work with Spotify Free?
Partly. Spotify Free works for search and building playlists, but it will not play audio through the agent. For actual playback you need Spotify Premium.
What chat apps can I control the DJ from?
Any surface Hermes supports, including Telegram, Discord, SMS, and the terminal. You type the same natural-language commands no matter where you are.
Which model should I use for the DJ profile?
A fast, cheap model is best because music commands are short and frequent. I use deepseek-v4-flash, which keeps the agent responsive without running up costs.
How reliable are the cron jobs?
Very. My main backfill job has completed 714 scheduled runs without a single failure, and a second sync job has logged 57 runs. Once it is configured, it just keeps going.
About Jason Pollak
Jason Pollak is a marketing strategist with over 10 years of experience building campaigns for entertainment brands, artists, and businesses across music, film, television, eCommerce, and B2B SaaS. As Director of Marketing at Young Money Entertainment, he grew Lil Wayne’s Facebook following from 10 million to 50 million and managed over 60 million followers across the roster. He also served as Paid Media Director at Horizon Media, launching major TV shows for History Channel, A&E, WWE, and Lifetime, and led film marketing for Utopia Distribution, generating over $10 million in revenue on a $200K media spend. Jason specializes in paid media, organic social strategy, email automation, SEO, content development, and AI-driven marketing systems. He holds a BA in English Literature from Binghamton University and a Masters in Media Studies from Brooklyn College. Learn more at jasonpollakmarketing.com.