The best Hermes Agent memory for a business is the built-in memory with an Obsidian vault added on top, because it is free, it stays on your machine and it lets every agent share what your business already knows.
Per-profile memory is my second pick for anyone with separate clients, and the built-in memory on its own is my third pick if you are only starting out.
Memory is the difference between an agent you brief every morning and an agent that already knows your offers, your clients and what you decided last week.
In this post I rank six memory options, show how a coach, an agency and a content creator would each use them, and cover what they cost and what to keep out.
The Top 3 Hermes Agent Memory Picks For A Business
- 🥇 Built-in memory plus an Obsidian vault is the best overall, because your business knowledge lives in files you own and any agent can read them.
- 🥈 Per-profile memory is the best for client work, because each client gets a separate memory that never mixes with another.
- 🥉 Built-in memory on its own is the best first step, because it is already running and costs nothing.
What Hermes Agent Memory Actually Is
Hermes Agent is the open-source AI agent from Nous Research, and it ships with a small memory that persists between sessions.
The official docs describe two files.
MEMORY.md holds the agent's own notes about your environment and what it has learned, and it is capped at 2,200 characters.
USER.md holds your profile, such as your preferences and communication style, and it is capped at 1,375 characters.
Together that is roughly 1,300 tokens, which is a few short paragraphs.
Hermes loads both files into every session automatically, so those facts are always in front of the model.
It also keeps every past conversation in a local database that it can search when it needs an older detail.
That design is clever, because the always-on part stays small and cheap.
It also means the built-in memory can't hold your whole business, which is why the vault matters.
How I Ranked These For Business Use
I ranked on four things that matter when money and clients are involved.
The first is whether I use it in my own business, and my top picks are what I run daily.
The second is cost, because a memory system you pay for every month has to earn it.
The third is where client data ends up.
The fourth is whether a non-technical owner could set it up in an afternoon.
I ranked the external providers from the official docs, not from long-term testing, so treat that part as a starting map.
🔥 Want my exact Best Hermes Agent Memory For Client Work setup, step by step? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom
The 6 Best Hermes Agent Memory Options For A Business
1. Built-In Memory Plus An Obsidian Vault
A vault is a folder of plain text notes that holds your clients, offers, processes, call notes and goals.
Hermes reads it before it works and writes to it after it finishes.
It is best for owners who want an agent that sounds like the business and remembers decisions.
It is free, because Obsidian is a free app and the notes are only text files.
2. Per-Profile Memory
A profile is a separate Hermes home with its own config, keys, memory, sessions and skills.
It is best for agencies and freelancers who serve more than one client.
It is free, and you create one with a single command.
3. Built-In Memory On Its Own
This is the two small files and session search with no extras.
It is best for a solo owner testing Hermes for the first week.
It is free and already switched on.
4. Holographic
Holographic is a local fact store that the docs recommend for local-only memory with no external dependencies.
It is best for a business that wants deeper recall but can't send data to a third party.
It is free.
5. Honcho
Honcho is an external provider from Plastic Labs that models you across sessions.
The docs say it is best for multi-agent systems with cross-session context.
It runs on Honcho Cloud at Honcho's pricing, or you can self-host it for free.
6. The Other Official Providers
The docs also name Mem0, Hindsight, Supermemory, OpenViking, ByteRover, RetainDB and Memori.
They are best for teams that already use one of those services or want a specific feature, such as Mem0's automatic fact extraction.
Most have a free local or self-hosted mode, and RetainDB is listed at $20 a month.
Best Hermes Agent Memory Options Compared For Business
| Rank | Option | Best business use | Monthly cost | Where client data lives |
|---|---|---|---|---|
| 1 | Built-in memory plus an Obsidian vault | It suits a business brain that every agent shares. | It is free. | It stays in files on your machine. |
| 2 | Per-profile memory | It suits separate clients or companies. | It is free. | It stays in each profile's own folder. |
| 3 | Built-in memory only | It suits a first week with one agent. | It is free. | It stays in two local files. |
| 4 | Holographic | It suits private, local recall. | It is free. | It stays in a local database. |
| 5 | Honcho | It suits several agents that need to know the same user. | It is paid on the cloud and free when self-hosted. | It lives on Honcho Cloud or your own server. |
| 6 | Other official providers | It suits teams with a specific need. | It varies, and RetainDB is $20 a month. | It depends on the provider. |
🔥 Want the memory vault template I use in my own business? Inside the AI Profit Boardroom, I've got step-by-step videos on wiring Obsidian into Hermes, plus the installable Agent OS with the shared memory already connected. You also get four coaching calls a week with 3,400+ members. → Get access here
How A Coach Or Consultant Would Use Hermes Memory
A coach's most valuable asset is what was said on calls.
The move here is to drop every call note into the vault's raw folder and let Hermes distil the key points into short wiki pages.
In my Hermes second brain guide, I call call notes gold for exactly this reason.
Once thirty days of calls are in there, you can ask a question such as "What were the most common objections in the last 30 days?"
Hermes answers from your own notes and names the notes it used.
You can then ask it to build a training on handling those objections, because the raw material is already organised.
The built-in USER.md file handles the small stuff, such as the tone you like and the format you want answers in.
How An Agency Would Use Hermes Memory
An agency has a different problem, which is that client A must never leak into client B.
This is where profiles earn their place.
The profiles docs say each profile gets its own config, keys, memory, sessions and skills.
You would create one profile per client and give each one its own memory.
A session on one profile and a Telegram bot on another profile do not share notes.
I get asked inside the AI Profit Boardroom all the time how to set up an agent system when you have separate companies or clients.
My answer is one profile per client for the private memory, and one shared vault for how your agency works.
The shared vault holds your processes, your reporting format and your writing rules.
Each client profile holds that client's history, preferences and quirks.
There is one hard rule from the docs, which is that you should never point two agent processes at the same Hermes home.
Both would write memory automatically and compound each other's entries until nothing in there is something you chose.
🚀 Ready to put Best Hermes Agent Memory For Client Work to work in your business? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom
How A Content Creator Would Use Hermes Memory
A creator needs the agent to sound like them every single time.
In my own vault I keep a Writing and Voice folder with my style rules, my banned words and the structures that work.
Every agent reads that folder before it writes anything.
That one habit removes most of the "this sounds like AI" problem, because the agent isn't guessing my voice.
I also capture ideas by talking.
A voice note or a message marked as a note goes straight into the Inbox folder with no title, no tags and no sorting.
Sorting is the agent's job, and it happens when a session files itself at the end.
How A Small Team Would Share Hermes Memory
Teams always ask whether everyone can share one memory.
The docs say that when two agents need shared memory, you should give each its own profile and use an external memory provider.
Honcho is the provider the docs point to for multi-agent work, because it keeps one shared view of the user with a separate identity for each agent.
I solve the same problem with the vault instead.
Seven of my agents read the same folder, and each one has its own sub-folder to write in.
The rule that makes it work is that every agent appends and no agent wipes.
Without that rule, several agents writing to one file turn it into mush within about a week.
How To Set Up The Best Hermes Agent Memory In One Afternoon
You can build my number one pick in six steps.
First, install Hermes and have one normal conversation, so the built-in memory files exist.
My post on the best Hermes Agent setup covers the install.
Second, install Obsidian and create an empty vault.
Third, give Hermes the full folder path and ask it to create the folders, which in my case are Inbox, Daily, Projects, Areas, Resources, Memories, Archive, Wiki and Agentic OS.
Fourth, write an About You note with who you are, what you sell, your numbers and your voice.
Fifth, add a protocol file that tells every agent to read before it works and write before it leaves.
Sixth, add the rule that the agent must never dump the whole vault and must say so when the answer isn't there.
That last rule is what turns a confident guesser into something you can trust with client work.
When you add a new client later, run hermes profile create with the client's name to give that client a separate memory.
What A Business Should Store In Hermes Memory
The built-in files are tiny, so use them for facts that must be present in every session.
| Put this in memory | Keep this out of memory |
|---|---|
| Put in your preferred tone, format and reporting style. | Keep out passwords, API keys and login tokens. |
| Put in corrections you have made more than once. | Keep out card numbers and bank details. |
| Put in where your key folders and tools live. | Keep out raw exports, logs and long documents. |
| Put in project conventions and client preferences. | Keep out anything a contract says must not leave a client's systems. |
| Put in completed work with a date. | Keep out one-off details you'll never need again. |
Long material, such as call transcripts and process documents, belongs in the vault.
The official docs make the same split, because they tell the agent to skip raw data dumps and easily re-discovered facts.
The Privacy Rules I Follow With Client Memory
I learned this one on camera.
When I toured Herald OS, I opened its Memory page to show everything Hermes remembered, and my Telegram details were on screen.
That page is a list of true facts about you and your business.
So I now check memory before I screen-share, record a video or jump on a client call.
I also keep secrets out of the vault completely, and my own protocol file says "no secrets, API keys or tokens in the vault".
If you choose a cloud provider, your conversations are synced to that company's servers.
The docs say Hermes scrubs recognisable secrets before it hands anything to a provider, which is a good safety net.
It is still your job to read the provider's terms and to check what your client contracts allow.
For an agent that talks to other people, you can turn on write approval so nothing is saved until you say yes.
My tour of that interface is in the Herald OS post.
🔥 Want help building a memory system for your own clients? Inside the AI Profit Boardroom, you can share your screen on one of four coaching calls a week and get your profiles, vault and memory rules set up properly. You'll be alongside 3,400+ members building real automations. → Join the AI Profit Boardroom
What Hermes Agent Memory Can't Do For You
Memory will not fix a weak model.
The docs say small local models often claim they saved something without calling the memory tool.
If that happens, use a stronger model for the setup, and my post on the best Hermes Agent model helps you pick one.
Memory is also only as good as what you feed it.
A vault with three notes in it gives generic answers, and a vault with thirty days of real calls and decisions gives answers in your own words.
Only one external provider can be active at a time, so you can't stack Honcho and Mem0 together.
Related Reading
- The best Hermes Agent setup walks through the full install around this memory layer.
- The best Hermes Agent skills covers how Hermes saves repeatable processes.
- The Hermes Agent plugins marketplace explains where memory plugins come from.
Also On Our Network
- 🌐 The step-by-step Hermes memory setup for operators
- 🌐 Hermes memory options scored and ranked
- 🌐 How the memory layer fits into an Agent OS
- 🌐 The models that save memory reliably, on GoldieBench
FAQ: Best Hermes Agent Memory For Business
What is the best Hermes Agent memory for a business?
My pick is the built-in memory with an Obsidian vault on top.
The vault holds what your business knows, and the built-in files hold the short facts needed in every session.
How do I keep client memory separate in Hermes Agent?
Create one Hermes profile per client.
The docs say memory is scoped per profile by design, so each client's notes stay in their own folder.
Does Hermes Agent memory cost money?
The built-in memory, profiles, a vault and Holographic are all free.
Cloud providers set their own prices, and the docs list RetainDB at $20 a month.
Should I put client data in Hermes memory?
Keep passwords, API keys and payment details out of every memory store.
Keep client notes in local files, and read your contracts before you use a hosted provider.
Can my whole team share one Hermes memory?
Don't point two agents at the same Hermes home.
Use an external provider or a shared vault with an append-only rule instead.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I've built Goldie Agency into a seven-figure SEO and link building agency.
- I've grown my YouTube channel to 400,000+ subscribers.
- I run GoldieBench, where I score AI models on real one-shot builds.
- I wrote "Link Building Mastery", which is available on Amazon.
→ Get my best AI training inside the AI Profit Boardroom
Feed the vault, give each client a profile, keep secrets out, and your business will be running the best Hermes Agent memory I know how to build.
