Kali & LLM: Completely local with Ollama & 5ire

We are extending our LLM-driven Kali series, where natural language replaces manual command input. This time however, we are doing everything locally and offline. We are using our own hardware and not relying on any 3rd party services/SaaS.

Note: Local LLMs are hardware-hungry. The cost factor here is buying hardware and the running costs. If you have anything that you can re-use, great!

GPU (Nvidia)

Let’s first find out what our hardware is:

$ lspci | grep -i vga

07:00.0 VGA compatible controller: NVIDIA Corporation GP106 [GeForce GTX 1060 6GB] (rev a1)

$

NVIDIA GeForce GTX 1060 (6 GB).

Drivers

We will check that our hardware is ready by making sure “non-free” proprietary drivers are installed. The non-free option allows for CUDA support which the open-source, nouveau, drivers lack.
At the same time, make sure our Kernel and headers are at the latest version too:

$ sudo apt update

[...]

$

$ sudo apt install -y linux-image-$(dpkg --print-architecture) linux-headers-$(dpkg --print-architecture) nvidia-driver nvidia-smi

[...]

│ Conflicting nouveau kernel module loaded │

│ The free nouveau kernel module is currently loaded and conflicts with the non-free nvidia kernel module. │

│ The easiest way to fix this is to reboot the machine once the installation has finished. |

[...]

$

$ sudo reboot

Testing

Once the box is back up and we are logged in again, we can do a quick check with nvidia-smi:

$ lspci -s 07:00.0 -v | grep Kernel

Kernel driver in use: nvidia

Kernel modules: nvidia

$

$ lsmod | grep '^nouveau'

$

$ lsmod | grep '^nvidia'

nvidia_drm 126976 2

nvidia_modeset 1605632 3 nvidia_drm

nvidia 60710912 29 nvidia_drm,nvidia_modeset

$

$ nvidia-smi

Tue Jan 27 14:33:31 2026

+-----------------------------------------------------------------------------------------+

| NVIDIA-SMI 550.163.01 Driver Version: 550.163.01 CUDA Version: 12.4 |

|-----------------------------------------+------------------------+----------------------+

| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |

| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |

| | | MIG M. |

|=========================================+========================+======================|

| 0 NVIDIA GeForce GTX 1060 6GB Off | 00000000:07:00.0 On | N/A |

| 0% 30C P8 6W / 120W | 25MiB / 6144MiB | 0% Default |

| | | N/A |

+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+

| Processes: |

| GPU GI CI PID Type Process name GPU Memory |

| ID ID Usage |

|=========================================================================================|

| 0 N/A N/A 969 G /usr/lib/xorg/Xorg 21MiB |

+-----------------------------------------------------------------------------------------+

$

Everything looks to be in order.

Ollama

Next up, we need to install , or follow below for v0.15.2 (latest at the time of writing, 2026-01-27):

$ sudo apt install -y curl

[...]

$

$ curl --fail --location https://ollama.com/download/ollama-linux-amd64.tar.zst > /tmp/ollama-linux-amd64.tar.zst

[...]

$

$ file /tmp/ollama-linux-amd64.tar.zst

/tmp/ollama-linux-amd64.tar.zst: Zstandard compressed data (v0.8+), Dictionary ID: None

$ sha512sum /tmp/ollama-linux-amd64.tar.zst

1c16259de4898a694ac23e7d4a3038dc3aebbbb8247cf30a05f5c84f2bde573294e8e612f3a9d5042201ebfe148f5b7fe64acc50f5478d3453f62f85d44593a1 /tmp/ollama-linux-amd64.tar.zst

$

$ sudo tar x -v --zstd -C /usr -f /tmp/ollama-linux-amd64.tar.zst

[...]

$

$ sudo useradd -r -s /bin/false -U -m -d /usr/share/ollama ollama

$

$ sudo usermod -a -G ollama $(whoami)

$

$ cat <<EOF | sudo tee /etc/systemd/system/ollama.service >/dev/null

[Unit]

Description=Ollama Service

After=network-online.target

[Service]

ExecStart=/usr/bin/ollama serve

User=ollama

Group=ollama

Restart=always

RestartSec=3

Environment="PATH=\$PATH"

[Install]

WantedBy=multi-user.target

EOF

$

$ sudo systemctl daemon-reload

$

$ sudo systemctl enable --now ollama

Created symlink '/etc/systemd/system/multi-user.target.wants/ollama.service' → '/etc/systemd/system/ollama.service'.

$

$ systemctl status ollama

● ollama.service - Ollama Service

Loaded: loaded (/etc/systemd/system/ollama.service; enabled; preset: disabled)

Active: active (running) since Tue 2026-01-27 14:44:39 GMT; 18s ago

[...]

$

$ ollama -v

ollama version is 0.15.2

$

The service is reporting to be active and running (and nothing is off in the logs files).

LLM

Now we need an LLM for Ollama to run! There are a few places to find pre-generated LLMs:

• (aka HF)

Which models you might ask? Time to experiment!

• We need a model which has “Tools” support. We will explain later why this is important.


• Your hardware will dictate how complex of a model you can run. The hardware we are using has 6GB of VRAM , so we will need a model size which requires less.

We have chosen 3 to test:

• - ollama pull llama3.2:3b


• of Kali, which means there isn’t any pre-installed tools.
  
  

  Sticking once again to (5ire-0.15.3-x86_64.AppImage at the time of writing, 2026-01-27) and make a menu entry:

  
  
  CONSOLE Kopieren

  $ curl --fail --location https://github.com/nanbingxyz/5ire/releases/download/v0.15.3/5ire-0.15.3-x86_64.AppImage > 5ire-x86_64.AppImage
  
  [...]
  
  $
  
  $ file 5ire-x86_64.AppImage
  
  5ire-x86_64.AppImage: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.18, stripped
  
  $ sha512sum 5ire-x86_64.AppImage
  
  bdf665fc6636da240153d44629723cb311bba4068db21c607f05cc6e1e58bb2e45aa72363a979a2aa165cb08a12db7babb715ac58da448fc9cf0258b22a56707 5ire-x86_64.AppImage
  
  $
  
  $ sudo mkdir -pv /opt/5ire/
  
  mkdir: created directory '/opt/5ire/'
  
  $
  
  $ sudo mv -v 5ire-x86_64.AppImage /opt/5ire/5ire-x86_64.AppImage
  
  renamed '5ire-x86_64.AppImage' -> '/opt/5ire/5ire-x86_64.AppImage'
  
  $
  
  $ chmod -v 0755 /opt/5ire/5ire-x86_64.AppImage
  
  mode of '/opt/5ire/5ire-x86_64.AppImage' changed from 0664 (rw-rw-r--) to 0755 (rwxr-xr-x)
  
  $
  
  $ mkdir -pv ~/.local/share/applications/
  
  mkdir: created directory '/home/kali/.local/share/applications/'
  
  $
  
  $ cat <<EOF | sudo tee ~/.local/share/applications/5ire.desktop >/dev/null
  
  [Desktop Entry]
  
  Name=5ire
  
  Comment=5ire Desktop AI Assistant
  
  Exec=/opt/5ire/5ire-x86_64.AppImage
  
  Terminal=false
  
  Type=Application
  
  Categories=Utility;Development;
  
  StartupWMClass=5ire
  
  EOF
  
  $
  
  $ sudo ln -sfv /opt/5ire/5ire-x86_64.AppImage /usr/local/bin/5ire
  
  '/usr/local/bin/5ire' -> '/opt/5ire/5ire-x86_64.AppImage'
  
  $
  
  $ sudo apt install -y libfuse2t64
  
  [...]
  
  $
  
  

  
  

  We can now either use the menu, or call it from a terminal.

  
  

  ---

  
  

  Now we need to configure 5ire to use Ollama (for LLM) and mcp-kali-server (MCP server):

  
  

  Let’s now setup 5ire to use Ollama.

  
  

  
  
  

  
  
  

  ---

  
  

  Let’s toggle Default to Enable it

  
  

  
  
  

  
  
  

  
  
  

  
  
  

  ---

  
  

  Again, checking status:

  
  
  CONSOLE Kopieren

  $ ollama ps
  
  NAME ID SIZE PROCESSOR CONTEXT UNTIL
  
  qwen3:4b 359d7dd4bcda 3.5 GB 100% GPU 4096 2 minutes from now
  
  $
  
  

  
  

  ---

  
  

  
  
  

  
  
  

  ---

  
  

  Now to fill in the boxes:

  
  
  
  
  • Name: mcp-kali-server
  
  
  • Description: MCP Kali Server
  
  
  • Approval Policy: …Up to you
  
  
  • Command: /usr/bin/mcp-server
  
  
  
  

  Save

  
  

  
  
  

  
  
  

  ---

  
  

  We can see what we now have on offer. ... -> Browse

  
  

  
  
  

  
  
  

  Testing

  
  
  
  
  • New Chat -> Ollama
  
  
  
  

  
  

  Can you please do a port scan on scanme.nmap.org, looking for TCP 80,443,21,22?

  
  

  
  

  
  
  

  
  
  

  
  
  

  
  
  

  Wonderful!

  
  

  Recap

  
  

  As a recap:

  
  
  
  
  • On our Kali local instance, we enabled our GPU for development.
  
  
  • We setup Ollama and grabbed a few LLMs, such as qwen3:4b.
  
  
  • Setup a MCP server, MCP-Kali-Server.
  
  
  • We installed a GUI interface, 5ire.
  
  
  • We setup 5ire to use Ollama’s LLMs as well as MCP client to use mcp-kali-server.
  
  
  • We then used it all to-do a nmap port scan of scanme.nmap.org …all processed locally!
  
  
  
  

  We may be talking about AI, but AI was not used to write this!

  
  

  ---

  
  

  Find out more about advanced red teaming for AI environments at OffSec.com.

  Vollständiger Original-Artikel

  Den kompletten Beitrag mit allen Details direkt auf kali.org lesen.

  ↗ Original-Artikel auf kali.org lesen