The best Hermes Agent memory setup for most people is Hermes’s built-in memory plus a plain-text vault like Obsidian — then one memory provider if you need more. In the video below I show Hermes, Jarvis-style voice and Obsidian memory working together. Here’s how Hermes memory works, every provider compared, and what I’d choose.
Key takeaways
- Built-in memory (MEMORY.md and USER.md) is always on.
- Hermes supports nine external memory providers; only one can be active.
- Honcho models you; Hindsight builds a knowledge graph.
- Three providers leave core on 15 October 2026 — install them from the catalogue.
How Hermes Agent memory works
Out of the box, Hermes keeps two memory files — MEMORY.md for what it learns and USER.md for what it knows about you — and nudges itself to save useful facts. It can also search past sessions with full-text search and summarise what it finds. An external provider adds to this; it doesn’t replace it.
Best Hermes Agent memory providers compared
| Provider | Best for | Runs |
|---|---|---|
| Built-in (MEMORY.md + USER.md) | Everyone — always on | Local |
| Honcho | Modelling you as a user over time | Catalogue plugin |
| Hindsight | Knowledge-graph memory with a reflect tool | Catalogue plugin |
| Mem0 | Server-side fact extraction and semantic search | Catalogue plugin |
| Supermemory | Semantic long-term memory with profile recall | Catalogue plugin |
| OpenViking | Self-hosted context database with folder-style browsing | Catalogue plugin |
| Holographic | Local SQLite fact store, no external services | Local (leaving core 15 Oct 2026) |
| ByteRover | Local-first knowledge tree with optional cloud sync | Local CLI (leaving core 15 Oct 2026) |
| RetainDB | Cloud memory API with hybrid search | Cloud, API key (leaving core 15 Oct 2026) |
| Memori | Cloud memory with explicit recall tools | Cloud, external installer |
Install a catalogue provider with hermes plugins install <name> and run hermes memory setup. Check what’s active with hermes memory status, and switch the external provider off with hermes memory off.
Want this working in your business, not just bookmarked? Setting up memory that makes your agents smarter is exactly what we build together inside the AI Profit Boardroom — 3,400+ members, four live calls a week, and a full one-hour DeepSeek Harness course in the classroom.
Prefer it mapped 1-on-1 first? Book a free strategy session and we’ll plan it for your exact situation.
Join the AI Profit Boardroom →Book a Free Strategy Session →
My pick for the best Hermes Agent memory
- Everyone: built-in memory plus an Obsidian vault Hermes can read and write.
- Personal assistant use: add Honcho, which models your preferences over time.
- Research and knowledge work: add Hindsight for knowledge-graph memory.
- Fully local and private: Holographic or ByteRover — from the catalogue once they leave core.
Why a plain-text vault matters
Memory providers are great, but your most valuable knowledge should live in files you own. I keep mine in an Obsidian vault that every agent in my Agent OS can read, so nothing is locked inside one tool. If you ever switch providers, or agents, the vault comes with you.
Keep Hermes memory clean
- Review MEMORY.md and USER.md every week.
- Delete wrong or outdated facts.
- Use separate profiles or bots for separate jobs, so memories don’t blur.
- Don’t store passwords or keys in memory.
My Hermes memory management routine takes 15 minutes a week.
Built-in memory vs a provider
| Built-in memory | External provider | |
|---|---|---|
| Setup | None | One install command |
| Where it lives | Local files | Local or cloud, depending on provider |
| Best at | Facts and preferences | User modelling, knowledge graphs, semantic search |
| Cost | Free | Free to paid, depending on provider |
Most people should run built-in memory for a few weeks before adding a provider — you’ll know exactly what’s missing.
What changes on 15 October 2026
The Hermes docs say Holographic, RetainDB and ByteRover are scheduled to leave core on 15 October 2026. If you use one, check the docs after that date for how to keep using it, likely through the plugin catalogue like Honcho, Mem0 and the others.
How memory makes Hermes better over time
Memory is what turns Hermes from a clever chatbot into an assistant that knows you. It remembers your business, your audience, your style rules and the mistakes you’ve corrected, so every task starts from where the last one finished. Combined with skills, that compounding effect is huge: after a few weeks you stop re-explaining yourself and start getting work that fits first time. That’s why keeping memory accurate and well organised is one of the highest-value habits you can build with any AI agent.
There’s a privacy side too. Built-in memory and a local vault keep everything on your machine. Cloud providers can be more powerful, but they mean your memories live on someone else’s servers. Decide what you’re comfortable with before you install a provider, and keep anything sensitive out of memory entirely. For most people, a local-first setup with a well-organised vault gives you the best balance of power and privacy.
Finally, give it time. Memory gets more valuable the longer you use Hermes, so judge it after a month of real work, not after a single afternoon. By then you’ll see exactly where it helps and where a provider would add something worthwhile.
How I’d get started this week
Open MEMORY.md and USER.md today and read what Hermes has saved. Fix anything wrong, delete anything stale, and add three facts you want it to always know — your business, your audience and your preferred style. Connect an Obsidian vault if you use one. Only then decide whether you need an external provider, and if so, pick the one that matches how you use Hermes.
Memory mistakes to avoid
- Never reviewing memory — wrong facts compound.
- One profile for everything — research, content and personal memories blur.
- Stacking providers — only one external provider can be active.
- Saving secrets — keep keys and passwords out of memory.
- No backup — keep your vault in a folder that’s backed up.
Avoid those and memory becomes Hermes’s biggest advantage over a normal chatbot.
Memory and voice together
In the video above I combine Hermes memory with a Jarvis-style voice and an Obsidian vault. You talk, Hermes remembers, and the notes land in your vault where every agent can use them. That loop is what makes an agent feel like it actually knows you.
The bottom line on the best Hermes Agent memory
The best Hermes Agent memory is built-in memory, a vault you own, and at most one provider chosen for your use case. Keep it clean and Hermes gets smarter every week.
FAQ: best hermes agent memory
What is the best memory for Hermes Agent?
Built-in memory plus an Obsidian vault, with Honcho or Hindsight added if you need more.
How many memory providers does Hermes support?
Nine external providers, and only one can be active at a time alongside built-in memory.
How do I set up a Hermes memory provider?
Run hermes plugins install with the provider name, then hermes memory setup.
Which Hermes memory is fully local?
Built-in memory, Holographic and ByteRover run locally.
Does Hermes forget between sessions?
No — built-in memory and session search persist across sessions.
Two ways I can help from here. Join the AI Profit Boardroom — it’s where your Hermes memory setup gets built with 3,400+ members doing the same.
Or grab a free strategy session and bring your questions.
Join the AI Profit Boardroom →Book a Free Strategy Session →
About Julian Goldie
I’m Julian Goldie — founder of Goldie Agency, best-selling author, and an AI educator with 428K+ YouTube subscribers. I’ve spent 10+ years in SEO and online business and I test every AI agent on my own work before I recommend it. Join the AI Profit Boardroom for the daily builds, or book a free strategy session to talk through yours.
Related reading
Last updated October 2026. This page is a living guide to best hermes agent memory — the facts here move fast and it is updated as they do.
