OpenClaw + Obsidian gives you super powers

OpenClaw's memory sucks. In this video, I'm going to show you how to make it great. I figured out a system using

great. I figured out a system using Obsidian and OpenClaw that has made my open claws memory basically flawless. If

you stick with me until the end of the video, doesn't matter if you use OpenClaw or Hermes, your AI agent will get super powers. Let's get into it. So,

real quick, here is a preview of my system. I'm going to show you how to set

system. I'm going to show you how to set it up. I'm going to show you how it

it up. I'm going to show you how it works. It's going to be super easy,

works. It's going to be super easy, don't worry. This is Obsidian. Inside

don't worry. This is Obsidian. Inside

Obsidian, I have new workspaces for all my agents. So, I use Hermes, I use

my agents. So, I use Hermes, I use OpenGL, I'm both here. But I have some very interesting things going on here.

First, I have daily logs. So, every day, my agent is automatically creating daily logs that covers everything we talk about high level. Every important task,

every single thing we do goes into our daily log. We also have a lot of other

daily log. We also have a lot of other things going on here that's superpowering our memory. And again, I'm going to show you how to set this all up in a second. We have mistakes. So we

have a mistakes file. Anytime I say, "Hey, Open Claw, hey Hermes, f you. You

did this wrong. What's wrong with you?"

It goes and it logs that mistake in the mistakes file. So it knows what it did

mistakes file. So it knows what it did wrong and it knows how to improve. We

also have some things around working context to allow it to have kind of a dynamic file with whatever context it needs in there. And we also have another thing agent shared. This is a shared

workspace that every agent has access to. So, when one agent does something,

to. So, when one agent does something, all the other agents have access as well. I'm going to go through how all

well. I'm going to go through how all this works, how to set it up, all that in just a second. But real quick, if you don't know what Obsidian is, this is it right here. It is a completely free

right here. It is a completely free tool. I'll put a link for it down below.

tool. I'll put a link for it down below.

All it basically is is a markdown user interface. It is a way for you to create

interface. It is a way for you to create and organize markdown file. It's super

simple, super clean, super lightweight.

For those who don't know what a markdown file is, they're just simple text files, but you're able to kind of like HTML, put formatting in it. It's basically the language of AI agents. It's the easiest

way for AI agents to read and understand things. So markdown files have been used

things. So markdown files have been used basically in every AI agent that's been created over the last year. OpenClaw's

completely built off them. If you look in your OpenClaw folder, it's all markdown files. So markdown has kind of

markdown files. So markdown has kind of become the language of AI over the last couple years. So everything we're going

couple years. So everything we're going to be doing in here will be markdown files inside Obsidian, which is is just a super easy program for both humans and

AI agents to organize and read files. So

let's talk about how your memory in OpenClaw works and exactly what we're changing here to make it better.

Basically, layers 1, 2, and four already exist in your OpenClaw. What we're doing is adding a third layer right here. Your

layer one is your builtin memory. These

are like the most important things that it needs to know like its name, where the important files live on your computer. This is like the sticky note

computer. This is like the sticky note layer. This is like when you put a

layer. This is like when you put a sticky note on your monitor to like remember really simple important things.

This is like the quick here's your name, here's who you are, what you do. This is

the built-in memory. This is included in every single prompt. So every prompt you send includes this memory. Then you have agents.md and soul.md. agents.mmd is

agents.md and soul.md. agents.mmd is

your rules file. So all the important rules for your agent are loaded in agents.mmd soul.md file. This also lives

agents.mmd soul.md file. This also lives in your open claw folder. This is its personality. This is what describes it

personality. This is what describes it how to communicate with you and what makes it fun and interesting and whatever other rules you want specifically for its personality. You

also have a session search. So this is built in as well which basically just has a record of every session you've had, every cron job you've had. If your

open claw is starting to move kind of slow, it's probably because this fourth layer is getting overloaded. But we'll

discuss that in a new video. If you want another video on how to speed up your open claw, let me know down below. But

the layer we're adding here is layer three. This is going to be the Obsidian

three. This is going to be the Obsidian vault. This is going to add a whole new

vault. This is going to add a whole new layer of really important things for your Open Claw or Hermes agent or whatever AI agent you're using to remember. Unlike these other layers, it

remember. Unlike these other layers, it is not automatically injected into every prompt. Right? Session search, all your

prompt. Right? Session search, all your history of sessions injected with every prompt. The rules file, the soul file,

prompt. The rules file, the soul file, the built-in memory, every prompt you said includes that. But the slight tweak here is not everything in your vault is included. So you can put a ton of stuff

included. So you can put a ton of stuff in your Obsidian vault. Like everything

you talk about every single day. How

this works is your agent at the start of every session is going to check its Obsidian vault for anything it needs to know about that session. It's also going

to pull from it exactly when it needs it. So, it's not overloading your

it. So, it's not overloading your context. So, to give you an example why

context. So, to give you an example why it loads at the beginning of every session, one big problem with OpenClaw up to this point is when it has compaction, sometimes it gets stupid.

Sometimes it forgets things from 5 seconds ago when it has a compaction.

Now, what's going to happen with every compaction is your open claw is actually going to go into your obsidian vault and see what it talked about right before that compaction. That will make it so it

that compaction. That will make it so it really doesn't forget things on the compaction. You don't even notice the

compaction. You don't even notice the compaction. I'm going to be honest, I

compaction. I'm going to be honest, I haven't noticed a single compaction since putting in this system. It's also

memory on demand. What does that mean?

So, if you say something like, "Hey, let's work on that project from 3 days ago." it'll be able to in that moment

ago." it'll be able to in that moment look at your daily log from three days ago, see exactly what you're talking about and start working on that project again. For me, that was a huge issue

again. For me, that was a huge issue before this system is we'd work on this small little project. Maybe I'm like adding AI to my smart glasses and I work on it for like 5 minutes and I go and do

other things and get distracted. If I

come back a few days later, it'll forget about that small tiny project. But now

with our added context in our obsidian vault, I can say, "Hey, that thing from 3 months ago," and it'll go and find it very, very easily. So, this is really the layer that's going to fill all the

tiny gaps in memory that have pissed you off so much over the last couple months.

And what I love about this system, too, is it adds an agent shared workspace.

This is going to be memory that all your agents can look at and share together.

So, for instance, here's knowledge. I

was working on a YouTube video with my Hermes. We scripted it out. We worked on

Hermes. We scripted it out. We worked on the video together. We had it all set to go. And now because it was placed in my

go. And now because it was placed in my agent shared directory, now my open claw when I go over that, I can start resuming work on that video immediately.

I can say, "Hey, I was just working on a video with Hermes." I go, "Oh, I see it.

There it is. Let's work on it together."

So you're basically making an entire share directory, too. So now you're This is how all your agents connect, right?

Ideally, you have at least two AI agents. If you're working with just one

agents. If you're working with just one right now, just one OpenClaw or just one Hermes, I highly recommend expanding to at least two. You can use them side by side. They work really good side by side

side. They work really good side by side once you have that multi- aent system going. This agent shared folder is going

going. This agent shared folder is going to make it so easy for them to communicate, share context, and work on projects together. So, this guide I'm

projects together. So, this guide I'm about to give you, this prompt I'm about to give you is going to set up this entire system. One for your agents and

entire system. One for your agents and one for your shared workspace so they can work together. By the way, if you've learned anything so far, make sure to leave a like down below. Also, subscribe

and turn on notifications. All I do is make amazing videos about AI. I also

live stream three times a week. I'm also

going to do a live boot camp on Open Claw Memory next week as well in the Vibe Coding Academy. Make sure to join the number one community in AI. Link for

that is down below. Okay, so let's talk about setting it up. This is actually the easy part cuz I'm going to give you a prompt. I'm going to put it down

a prompt. I'm going to put it down below. You're going to copy and paste it

below. You're going to copy and paste it in and you'll have the system set up.

Obviously, first thing you want to do is install Obsidian. So, that is good to

install Obsidian. So, that is good to go. Then your agent can find it. But

go. Then your agent can find it. But

here is the prompt. I put this in the Vibe Code Academy a few weeks ago. I'll

put it down below so you can just copy and paste it. It's going to basically describe to your agent the four layer system so it knows exactly how it works.

And it's also going to go through the flow as well. So, it's going to talk about to your agent what the flow is, how to write these memories, how to look into the memories, what to do when a compaction hits, everything like that.

One part here you want to customize in here. Uh, it says location iCloud

here. Uh, it says location iCloud Obsidian Vault shared with OpenClaw. If

you're just installing Obsidian now, just get you can get rid of the iCloud part and just say the Obsidian Vault.

This might confuse your OpenClaw agent if it doesn't know where the Obsidian vault is. So, you could, if you want to

vault is. So, you could, if you want to make this easier for your agent, just find the location of the vault you create for Obsidian. That's the location basically of your Obsidian folder when you install it. And you can just put the

path for it right there, and you'll be good to go. This is all down below. Copy

and paste it, give it to your agent. You

hit enter on that. It's going to start building the system for you. It's going

to put into its own rules file how to use this system. The one risk you might have that you just want to take care of is validate after you get this set up and it's actually storing memories to

your vault. Sometimes if you're on some

your vault. Sometimes if you're on some models that aren't as good with taking action immediately or writing rules immediately, it won't write all the rules for itself to use the vault. So

make sure right after you do this, check to see if memories are being written there. If they're not, say, "Hey, that

there. If they're not, say, "Hey, that rule system we just created, make sure you burn into your agents.mmd file how to use this system." So, moving forward, you use it with everything we do. If you

do that, it should be good to go. It

should be in the rules file. Your agent

will be using this memory system moving forward. A lot of other really cool

forward. A lot of other really cool things you can do here as well. Once you

have this memory system set up, you basically can use it to power other personal appears. So for instance, I

personal appears. So for instance, I have this Alex Finn wiki which is basically a place all my memories get stored that I can visually explore. So

instead of having to go into Obsidian, I can go in here. I can see what people are stored in my memories, what concepts are stored in my memories, what sources are stored in my memories, things like

that, and be able to easily explore things I've talked about in the past. So

you can have things like your own personal wiki. All I did for this was I

personal wiki. All I did for this was I went to my open client and said, "Hey, build a visual interface for our memories." almost like a personal wiki

memories." almost like a personal wiki and it went and built this. So, so many other really, really cool things you can do once you start building this personal

memory system inside of Obsidian. The

possibilities really are endless. A lot

more coming soon. So, make sure to subscribe and turn notifications. I drop

a lot of these prompts and stuff as I come up with them inside the Vibe Coding Academy. So, make sure to sign up for

Academy. So, make sure to sign up for that if you're interested. Also, let me know down below what you want me to talk about next. You want me to go more into

about next. You want me to go more into hardware and local models when it comes to these agents? Do you want me to talk about more about use cases and workflows or more about improving your agent like we just did in this video? Let me know

down below. I'm very interested. All my

down below. I'm very interested. All my

videos are based on your feedback. I

hope this was helpful. I'll see you in the next video.