---
title: 'Local OpenClaw & Ollama in 27 minutes'
source: 'https://youtube.com/watch?v=n2a1FfqjHcU'
video_id: 'n2a1FfqjHcU'
date: 2026-06-16
duration_sec: 0
---

# Local OpenClaw & Ollama in 27 minutes

> Source: [Local OpenClaw & Ollama in 27 minutes](https://youtube.com/watch?v=n2a1FfqjHcU)

## Summary

The video demonstrates how to run OpenClaw entirely locally using Ollama, avoiding cloud API costs and ensuring data privacy. It covers both a beginner setup (everything on one machine) and an advanced two-machine configuration with a Jetson Nano and an old gaming laptop. The creator explains how to select suitable local LLM models and configure networking for remote access.

### Key Points

- **Why run OpenClaw locally** [0:31] — Cloud costs ($100–$200), privacy concerns, service outages (Claw servers down), and bans on pro plan use for OpenClaw (Gemini, Claude) are key motivations.
- **Local vs cloud vs hybrid setup** [2:09] — Three options: fully cloud (OpenClaw hosted on AWS calling OpenAI/Claude), fully local (OpenClaw + LLM on same machine), or hybrid (local device calling cloud LLM).
- **Installing Ollama and a model** [4:23] — Go to ollama.com, run the terminal command to install Ollama, then download a model (e.g., Qwen 3.5 9B) using 'ollama run <model_name>'.
- **Installing OpenClaw via Ollama** [5:49] — Use 'ollama launch openclaw' to install OpenClaw directly. Choose a model (e.g., Qwen 3.5 9B) and enable no-think mode if reasoning is slow.
- **Adding Ollama models to existing OpenClaw** [7:38] — Use vibe coding (Claude Code/OpenAI Codex) to automatically update config, or ask OpenClaw via chat, or manually edit the config file. Restart gateway with 'openclaw gateway restart'.
- **How to choose the right LLM model** [10:48] — Download LM Studio to test models; pick ones with tool-use compatibility for agent use. Balance speed vs quality — Qwen 3.5 9B is current best balance for this setup.
- **Advanced two-machine setup** [14:00] — Jetson Nano (or Raspberry Pi) runs OpenClaw 24/7 on low power; old gaming laptop runs Ollama LLM server. Laptop is set to wake-on-LAN and hibernate when idle.
- **Configuring remote Ollama server** [16:55] — Find IP with 'ipconfig', run 'ollama serve' with host set to that IP, set static IP via router reservation, enable wake-on-LAN in BIOS and Windows power management.
- **Connecting OpenClaw to remote Ollama server** [23:56] — Use curl to test connectivity, then modify OpenClaw config (via vibe coding, chat, or manual edit) to point to remote server IP and model. Restart gateway.

### Conclusion

Running OpenClaw locally gives full control and privacy but requires setup effort and hardware. Beginners may prefer a cloud VPS, but once configured, a local AI infrastructure runs reliably and independently of cloud providers.

## Transcript

This tiny computer runs my Open Claw
locally 24 hours a day. No cloud APIs,
no token costs, and even if the internet
goes down, it's still working. But
getting this system working was way
harder than I thought. I had to test
different local models, configure
networking, and even split the system on
two machines. In this video, I'm going
to show you how to run Local Claw
locally, how to pick the right models
for your computer, and a setup I'm using
at home with a Jetson Nano and an old
gaming laptop. Why should you run Open
Claw locally? First of all, Open Claw is
expensive to run. It eats up a lot of
tokens and before you know it, you run
out of credits or you're spending $100,
$200 on your Open Claw. Secondly, if
you're concerned with privacy, you don't
want to send all your data to a public
LLM.
Everything stays within your network and
nobody else can see your data. Most
importantly, what I found was that my
open claw keeps going down because
either I'm out of credits or claw just
went down the other day. The servers all
went down and I couldn't call anymore.
So when your cloud provider is not
working for whatever reason, it's
updated as model. You always have
something running as long as your local
server is running. So your open claw is
available to you at all times. Another
thing that caught me by surprise is that
when I first started using Open Claw, I
was able to use it with my OpenAI
subscription, my Claude code
subscription, and my Gemini
subscription. And now Gemini and Claude
bans users who use the pro plan with
their open claw. So now you don't have
to worry about the policies of all the
different AI providers. You have full
control. But I must also say that the
concept of local AI is really good. But
getting it running locally, it's really
hard. Complex setup, you have to
understand networking configurations.
You have to have the hardware. You need
to have a fast computer. Otherwise, it's
just super slow. And it's taken me more
time to fix and configure things than I
would like. Running Open Claw locally
gives you a lot of freedom, but you also
become the system administrator. But
what does running open claw locally
mean? Actually, I think there's two
components. Number one is where does
your open claw run? And then open claw
needs to call an AI model to process the
request. And where is that AI model
running? You can have a cloud setup
where your open claw is hosted on a
server somewhere on Amazon on hosting
wherever it is and then that open claw
calls open AAI or claude. That's a fully
cloud setup or you can have a fully
local setup where you have a machine and
it runs open claw and on the machine
have a local llm model running so it can
provide the responses to your open claw
or you can have a hybrid setup where you
buy a Mac mini and it's running openclaw
right in front of you in your house but
it's calling open AI cloud gemini and
that openclaw calls a cloud LLM. So,
we're going to cover how to run OpenClaw
locally on your own device and hosting
the AI model locally within your house.
We're going to go through two examples
in this video. The first one is the
beginner setup, which is everything on
one machine. I'm going to show you how
to install OAMA and then using OAMA,
install Open Claw, and then run a local
model. So, that's the beginner setup and
then we're going to go [clears throat]
into my current setup. I have a Jetson
Nano which is like a tiny computer
almost like a Raspberry Pi running open
claw and then I have an old gaming
laptop that's running Lama and serving
the AI model. Why? Because number one, I
don't want to run open claw on my
MacBook because of security reasons. And
number two, my MacBook is really slow.
So when I talk to Open Claw, it takes a
long time for it to respond. So, by
running it on a old gaming laptop, I get
much better performance. And I'm going
to go into how to set that up. And I
also want to show how you don't have to
buy a super powerful computer like a Mac
Mini to do this. You can do this on some
computer you have lying around, an old
gaming laptop, and put them together and
make your own local Open Claw setup.
Okay, enough talking. Now, let me show
you step by step how to set up Open Claw
on your own computer. The first thing
we're going to do is go to alama.com and
install Alama. There are two ways to do
it. Either you can run this terminal
command or you can download a lama. So
you can press download and install it
there. But the best way to use a lama is
through the terminal. I'm going to copy
this command and then just run the
command. And then I can run a lama by
typing in a lama. And now I can run a
model launch cloud code launch codeex
launch open claw. So, let's start with
running a model. And then you're allowed
to choose different models. And it's
giving me recommendations based on my
specs. The recommended list is not the
best. So, you can choose, you know, GLM
4.7 Flash if you wish just to get
started. I already have a model
downloader, so I'm going to use that.
And I'm going to show you how to pick
the best one and update that later. And
I'm going to give it a test. Hi. All
right. And so, Alama is running. And now
let's download Quen 3.5. So I go to a
llama. I click on models. I click on
Quen 3.5 and I'm going to select Quen
3.59B latest. Copy this. And then I type
in OAMA run and then paste Quen 3.59B.
And it's going to start downloading the
model. And let me run that model. Okay.
And it's done. So let's give it a test.
Hi. And it works. The next thing we're
going to do is a new thing that open
claw has enabled which is you can now
use lama to install open claw. To
install open claw with lama all you need
to do is copy this command. Alama launch
open claw. Copy it. Then go to your
terminal and paste that in. And then
it's telling me to choose my model. I'm
going to choose quen 3.59b. I understand
the risk. Okay. So, it's finished
installing and I just sent it a message
saying, you know, I'm Keith and it
responded, but it's taking a very long
time. So, the problem with Quen 3.59B is
that it's got reasoning and it thinks a
lot before it does that. So, I'm going
to tell Open Claw to set it to no think
mode. It's taking too long to respond.
Okay, now it's set it to no thinking
mode. So, it should be faster. Now, now
that you're set up, you also want to
make sure your web interface is working.
So when you first installed it, you
should have an address like 127.0.0.1
18789. So let's go to that. We've opened
our browser and then we're going to
paste in the address. Now it's going to
say gateway token missing. When you
first install it, it should display a
URL with a token equals something. And
in my case I need to come to overview go
to open gateway token and my gateway
token is a llama. I press connect click
refresh and once I click refresh
I can see that on the top right the
health is okay and I'm connected is all
green. You can click on overview and see
stat is okay. Then if I come to chat,
you're going to see that the messages
that I've been sending earlier are
working. And I can also chat here. So
let's give it a try. And it's responded
to my high. It's working.
Congratulations. You have local llm
working with your open claw. I know that
a lot of people already installed open
claw. And if you're not using lama to
install open claw, how do you add lama
to your existing open claw? Well, if I
come here in my web dashboard, I come to
config and then I click on raw, you're
going to see the configuration file. And
in the configuration file, you can
change your configuration file to
models, providers, or llama. And then it
sets it to quen 3.5 9b. But I have to be
honest with you, I hate changing the
config file. It's really hard. You keep
making mistakes and it doesn't work. So,
what's the best way to do it? I'm going
to show you two ways to do it. Number
one, we're going to use any vibe coding
tool you have, OpenAI codeex or Claude
Code or Gemini, whatever you have, you
can use that to add to your model list.
And then number two is to directly tell
Open Claw to update your config file to
include Lama in your model selection.
So, let's go back to our terminal
interface. And if I select slash
open model picker, it's going to allow
me to search. I'm going to type in a
lama. Right now, it's only got a llama
quen 3.5 9b. Now, let's say it's not
even there. How do I add new models to
that? So, I'm going to exit this. And
you can use whatever you like. I'm going
to use claude. And I'm going to bring up
our llama. You can see I have GMA 3 4B
which I downloaded a long time ago and
it's not available in my model list. So
let's add that. I'm using open claw menu
and new addama
3 to my config get as a model. So what
it's going to do is that it's going to
search all the files on my computer find
the config file and it's going to add to
my config file so then they can find my
model and it's done. So, it's found the
config file and it's added GMA 4B
automatically without me manually going
in and making a mistake. Let's go open
model picker. Okay. And you can see that
now it's added GMA 4B. And before this,
I realized I needed to restart my
gateway for it to recognize. So, what
you do is you need to type in open claw
gateway restart and it will refresh and
reload config. and then you'll be able
to find the new model you just added. So
that's the easiest way to add your lama
models to an existing open claw when you
haven't installed it using. And the
other way is if you're already connected
to a cloud platform and you can chat
with open claw. What you can do is you
can say can you add gamma 3 to my config
file so I can choose as a model. Now
that will work too but sometimes I
realize it does funky things. The best
way is to use Claude Codeex or Gemini,
whatever AI coding agent you have to do
it, but this also works too. So you can
try that. And then the last way is to
just go in into your config file and
modify the configuration file manually.
Now that we have open claw installed,
this is the hardest part, choosing the
right LLM for Lama. And the trick is you
have to download another software
because in Olama it doesn't give you
much information. So download LM Studio,
try out different model and then once
you find the best model, run it in O
Lama because Lama works better with Open
Claw. It's got a smaller footprint and
it just runs better. So go to
lmstudio.ai and then download the app.
And once you have it, click on the
search icon, click on the discover icon,
and you'll see all this great
information. Compared to Lama, which
only gives you the model name, LM Studio
gives you a lot more. Most importantly,
it gives you a best match. So, it
recognizes your computer's
specifications and then recommends the
best one for you. What you have to pay
attention to is two things. Number one,
how big is the model? So, the larger the
model size, the more powerful it is, but
also the longer it takes to run. So,
it's recommending a lot of smaller
models for me, like 9B, 1.2 billion. But
let's look at the most downloads. You
have ones are 20 billion, 30 billion,
and that's just too much for my computer
to run. The second thing you should look
for is you probably want to look for a
model that is compatible with tool use
because it means it's designed for agent
use like Open Claw. So, pick models that
have this symbol on it and then just
click download. And after you've
selected it, you need to give it a test
run to test it speed and its
performance. So, let's give it a try.
I'm going to load a model and right now
I have Quen 4B
and I'm going to type test. Okay, so it
came back and you can see that this took
about 10 seconds and for me this is a
little bit slow for me. I want something
faster. So I'll test a different model.
So keep testing to find the one where
the speed is good for you and also
you're happy with the results it's
giving you. All right, I've been testing
a lot of models and I made another video
on setting up your own local AI, but
here's what I found. It's basically a
balance between speed and performance.
So, with Open Claw, a lot of people say
Kimmy K 2.5 is really good. It is really
good. The results that it comes back
with is really good, but my computer is
not fast enough to run it. So, I get
super slow response speeds. Then I tried
using LFM2, which is super fast and
lightweight. is one of the smallest and
fastest models out there. But the
results that it gives me are not very
smart and I've switched to the current
winner is Quen 3.59B. In terms of speed
and performance is the best one so far
on the market. But the key is you have
to keep constantly updating. I was using
Quen 3 6 months ago and now the model
has improved so much. So, the key is to
check for new models every month or two,
play around with it in LM Studio, and
then strike a balance between speed and
quality. And I'm excited because the
open- source models you can download are
getting much better very quickly. Okay,
now we're going to go into a more
advanced setup. And this is my setup. I
have a Jetson Nano, kind of like a
Raspberry Pi, running my Open Claw. But
although it is a cheap device, it's not
powerful enough to run AI models on it.
So I can only run open claw. So I use my
old gaming laptop and on that I install
llama and I run my AI model on that. Why
don't I just run open claw on my gaming
laptop as well? Well, because my gaming
laptop is not designed to be run 24/7.
And I do want it to run 24/7 because I
want to message it at any time and I
want it to run tasks overnight. is not
designed for that. The Jetson Nano is.
So, I have both computers running at
home and this only turns on when I need
to run AI models on it. And why the
Jetson Nano? Well, I wouldn't recommend
it. It's just something that I have
lying around. The best option is
actually a Raspberry Pi. So, you can get
an $80, $150 option out there where you
can buy a Raspberry Pi, buy a nice
little case, and you can leave it on
running with low electricity costs 24/7,
no problem. or you can get a Mac Mini.
More expensive, but you can even run
your LL models allin-one. So, it depends
on your budget. Enough talk. I'm going
to show you my setup with a Jetson Nano
running Open Claw and having my old
gaming laptop as a Lama LLM server
connecting to each other. Here's my
Jetson Nano. I bought a little case and
it even shows the temperature, the CPU
usage, the RAM.
That's my Jessen Nano running 24/7.
Here's my old gaming laptop. It's
running. It's on. And the screen isn't
even on to conserve energy. And I'm
plugged it in so that it can wake on
LAN. I'm on my Windows machine right
now, which will act as a server to my
JSON Nano. And I've gone to the website.
I've copied the command, which is this
command. And now I'm installing it on my
Windows machine. So, I typed in
terminal, open terminal, and now it's
installing OAMA. And so, I've installed
it. And the next thing you need to do is
to download a model. So, all you need to
do is put in this command, Alama run,
and then copy the model name and put it
in. So, you can go to models. We're
running Quen 3.59B. Copy that and then
paste it in and then hit enter. And it
will download the model and also run it
at the same time. Okay. And it's done.
So, let's give it test. And it's work.
Okay. And it's working. So, I'm going to
just exit this for a little sec. I'm
going to exit this. Now, this is only
running on my Windows, but in order for
it to be a server, I need to allow other
computers to access my Windows. So,
first of all, I need to type in IP
config. And you'll see that my IP
address is 1 192.168.68.62.
So I need to remember this address.
Going to copy it. Then you have to run a
command calledama serve. Now you'll see
that right now it's serving on what's
called localhost 11434,
port 11434. And even though this is on,
it doesn't mean that my other devices on
the network can access it. So, I'll need
to set the IP to the IP we just
discovered earlier. So, I'm going to
come out of here and I'm need to type in
this command dollar sign envah host
equals and then we're going to put in
the address we did earlier. Okay, I put
in the address and then now I type in
serve and the host is now at this
address. So from my other devices, my
Justin Ano, I can now call this address
and it should be able to communicate
with it and use this as my server for my
open call. But every time you turn off
the computer and you restart it, your IP
address may change. So I'm going to show
you how I set a static IP on my router.
And there's more. I'm gonna put this on
wake on land. So then my computer can
turn off when it's not being used and
then only when it's being called by my
open call, it'll wake up, turn on the
power and then run the llama. Every time
my gaming laptop shuts down and then
powers on again, it will get a new IP
address. And I don't want that because
if my Justin Nano is calling it, we want
a fixed address so it knows exactly
where it is. So depending on your router
settings, this is my router. I use TP
link. What I do is I can come to more. I
come to advanced.
I go to address reservation.
And basically I've reserved my device to
always have a static IP on 192.168
6865.
And so it's fixed and that's reserved
for my Alama server running on my gaming
laptop. So depending on your router
settings, you need to set a static IP.
And that's how you set it. Now I'm going
to show you how to set some settings. So
then you can have the computer turn off
and then wake on land. So it only turns
on when you actually need to use it. I'm
going to go and restart my
Windows.
And then while it's restarting, just
keep tapping the delete button to bring
up the BIOS. Okay, now the BIOS is
brought up. You want to go to advanced.
You want to enable USB power and sleep
in hibernation. You want to disable
fast boot and then you want to enable
wake on land and then you save and
reset. Now depending on your machine,
you might have a different BIOS. So
check with AI what settings you need,
but generally that's what you need to
enable wake on land. One more thing,
just make sure you use an Ethernet cable
so you have a cable plugged in because
it doesn't wake up if it's just on
Wi-Fi. Now that I have Alama as a
server, there's a quick little tip. By
default, the context length might be too
small. And so, you need to go and type
in a command that says show and then
quen 3.59B and it will show you your
context length. So right now it's at
26,000 2,144.
So I've increased it, but you might just
see 4,000 here. And if that's the case,
you need to increase it to a
recommendation of at least 16,000. The
way to set the context length is to
enter run your model name dash context
and then enter your number. Play around
with the number. Obviously, the larger
the context length, the better. But
also, if you have a really long request,
it might crash your computer. So, play
around with this number. I've set mine
to 26,000. At the minimum you should set
it to is around 16,000. Another thing
that I did was that once I've got the
llama running, I don't want my gaming
computer to be on all the time because
it's not designed that way and it'll
overheat really quickly. So, what I've
done is I've gone into my system power
and battery. I'm going to turn off my
screen after 5 minutes of nothing going
on. And then it will hibernate after 10
minutes. And if it goes to sleep, that's
fine. We're going to wake on land. So
then when my other devices call this
device, it will wake up. So how do you
do that? Well, you have to come to your
device manager and you have to look for
your internet connection devices. So I
have my Wi-Fi and I also have a LAN
connection where you plug in a cable. So
this is my Wi-Fi.
So if I want to wake my Wi-Fi, I can
come to power management and allow this
device to wake the computer and press
okay. Enable that. And the other one is
my LAN cable and I do the same. And you
can find this uh inside your network
adapter. And once you do that, the last
remaining thing you can do is to restart
your computer and in your BIOS set wake
on land. And one final thing is that you
don't have to set a static IP. There's
an even more advanced way to set this up
and that's by using tail scale. By using
tails scale, not only is it more secure,
but you can also access this server from
any device even when you are outside of
your house. In my current setup,
everything has to happen in the same
network and you have to be on the same
Wi-Fi to work. But let's say I'm out and
about and I want to access my Olama
server. I can do that using tail scale
and tunnel in and it's even more secure,
but that's for another video. So stay
tuned for my other video. I'm back on my
Mac. I have open claw installed. And
then we have a lama set up on my old
gaming laptop on a separate machine. So
the first thing we need to do is to
check if it can connect to it. You can
do that by using a curl command in your
terminal. So open your terminal, type in
this command, curl http your IP address
of your gaming machine/ ai/tags and it's
returned that it's got a model 3.59b. So
it's working. Okay, now that it's
connected to it, let's see if it
actually runs. So the next command is
curl API generate d choose a model. Say
my prompt is hello and I want to choose
stream. So I'll include this prompt
inside the description and you can see
that it's coming back with some response
and it's thinking. It's thinking. Let's
check back in when it gives me a result.
Okay. And it's done and it's giving me a
response. So it's working. So now we're
ready to go to open claw. So let's go to
open claw tui. So now that my MacBook
can connect to the server. There's three
ways to change the configuration
settings. So then open claw can use the
model remotely. Number one, you can go
into config file and paste this into it
and set your provider as a llama, your
IP address and the model. You can do
that manually, but every time I've done
that, it has not been successful. But if
you really want to do it, all you need
to do is go to finder, go to your home
directory, and then go command, shift,
and period. And you're going to see all
the hidden files. And in there, I'm
going to find open claw. So let's find
open claw openclaw.json. Let's open it.
And here I can input my I can copy this
the model provider and paste it into
here. But so far every time I've done
it, it has not been successful. So I'm
going to do in my opinion the best
option which is to use vibe coding to do
it. So in this case I'm going to use
cloud code. You can use openai Gemini.
I'm going to choose a folder and I'm
going to go back into the same folder.
I'm going to go back into open claw.
Okay. And then I'm just going to say
modify the open claw config file. So
then it points to my remote or llama
server. And then you plug in the
details, your IP address and the model.
And then it should automatically make
those changes. Okay, it says it's done.
And so I need to come back. And then the
thing I need to do is open call gateway
restart. So, I need to restart the
gateway to activate the change and then
let's test it out. And there we have it.
It's returned a response. Hi there, I'm
still awake. Let's start with the
basics. So, it's working right now. And
the third option is a chicken and egg
problem. You can ask open claw to update
the model similar to how I did it in
cloud code. Just tell it that you have a
new remote server. Give it the IP and
the model and it will update it. But the
chicken egg problem is that you need an
AI LLM provider connected to do this. So
either connect to your cloud service to
do this or you can run a really small
model on your local computer just so
then you can get the configuration done
so then openclaw can set its own config
file. So those are the three options you
can use to connect and change your model
to point to a remote server. So that's
how you run open claw locally. We
started with the beginner setup where
you're running openclaw on just one
machine and then I showed you my current
setup where a small computer runs open
claw 24/7 and another machine runs AI
model. But the bigger question is should
you run open claw locally at all? If
you're an absolute beginner honestly the
best option is to run open claw on a
virtual private server. So, you rent a
server somewhere and then call a cloud
model like OpenAI, Gemini, or Claude
using their API. There are ton of
one-click services out there for Open
Claw and you'll spend much less time
configuring things. But if privacy
matters to you or you're worried about
the API cost, then running Open Claw
could be a very powerful option. Your
data stays on your machine and your
system keeps running even if the cloud
goes down. And the number one
requirement is that you need a computer
that's fast enough to run a local LLM.
In my case, I'm just using an old gaming
laptop, and that works surprisingly
well. The trade-off is that local setups
require more work, networking, hardware,
testing different local LLM models until
you find the right balance between speed
and quality. But once everything is
running, you basically have your own AI
infrastructure at home. And that's
pretty sweet. Local AI gives you the
control, but it also makes you the
system administrator. If you've liked
this video, please like and subscribe to
my channel. If you want to learn more
about AI, you can join my free AI
community in the description. I look
forward to seeing you