Welcome to SubnetSlinger
Network Cowboys
The modern network engineer's toolkit, now on Linux. New here? This page walks you through it, step by step. (Prefer plain text? The complete reference is in README.txt.)
Get runningInstall - about a minute
SubnetSlinger doesn't need a system-wide install - you run it right out of the folder you downloaded. Here's the whole thing:
.tar.gz and choose Extract (or use the command below). Then open the new folder and, in most file managers, right-click an empty spot and choose Open Terminal Here - or just open your terminal app and cd into the extracted folder.chmod just marks it runnable (some extract tools drop that bit). It bundles its own runtime, so there's no .NET or framework to install first. To keep it around, move the folder anywhere you like (e.g. ~/Applications) - it runs from wherever it lives.dialout group (then log out/in) to open /dev/ttyUSB*.What you getWhat's in the box
Nearly 50 network tools in one app. A quick tour by what you'd reach for:
Diagnose & measure
Ping, Traceroute, Path ping/map, Port scanner, DNS, Speed test, WHOIS, SSL/TLS & HTTP inspectors
Discover & map
LAN Discovery, ARP table, MAC/OUI lookup, L2 neighbors (LLDP), Topology, SNMP, Wi-Fi scan
Connect & configure
SSH terminal, Serial console, SFTP, Multi-host, Config Builder, Provisioning, Drift detection, Inventory
Serve & capture
Local test servers (Syslog, TFTP, DHCP, DNS, RADIUS, TACACS+, SNMP trap) and live Packet Capture
Secure & assist
Vault for credentials, the Deputy AI + Posse multi-agent, and an MCP client for your own tools
Free forever - 9 tools
Ping, Traceroute, Path ping, Port scanner, DNS, Subnet calculator, ARP, MAC/OUI, Discovery. The rest unlock with the trial or Pro.
PrivacyWhat stays on your machine, what leaves it
Short answer: everything stays, unless you turn something on.
Stays local, always
- Every tool result
- Vault credentials (encrypted at rest)
- Metrics history, incidents, inventory
- Deputy AI chats when using local Ollama
- Anything you Save / Export to disk
Leaves only when you turn it on
- Cloud AI providers (opt-in, needs your key)
- Account sign-in - an email code to unlock Pro
- License purchase - Stripe in your browser
- Internet tools you invoke (speed test, WHOIS, public IP)
SpecsWill it run well on my machine?
The app is light. The only thing that wants a bigger machine is running the AI locally (Ollama) - and that depends on where the model runs, not just whether you use AI.
| How you run the AI | Your machine needs |
|---|---|
| No AI | Minimum |
| Cloud AI (your key) | Minimum + internet |
| A model on your own server | Minimum + reach it |
| Local Ollama (this box) | 32 GB RAM + a GPU |
| Minimum | Recommended | |
|---|---|---|
| CPU | 64-bit dual-core | quad-core or better |
| RAM | 8 GB | 16 GB (32 GB for local AI) |
| Disk | ~500 MB free | SSD, 2 GB+ (plus 5-10 GB per local model) |
| GPU | none (integrated OK) | local AI: 8 GB+ VRAM |
| Display | 1280x800 or larger | same |
Cloud AI, or any model you host on another box, keep this machine at the minimum. Only running the model on this box raises the spec. Full details are in README.txt.
Install helpersAdd-ons - only for the feature you use
SubnetSlinger doesn't bundle these. It builds the command and runs the one you have installed - and tells you at runtime if one isn't there. Click a row for the exact install command. Commands show apt; on Fedora use dnf, on Arch pacman.
Live packet Capture libpcap
Need it if: you'll use the Capture tool to record and inspect live network traffic.
The Capture tool wraps libpcap (usually already installed). If it's missing:
Then grant capture once with sudo ./install-linux.sh (step 4 above).
Command-line tools you drive (nmap, dig, mtr, tcpdump) CLI wrappers
Need it if: you'll drive nmap scans, dig lookups, mtr, or tcpdump from the CLI Wrappers tool.
Install only the ones you'll use:
SFTP / SCP receive (Local Servers) openssh-server
Need it if: you want devices (or people) to send files TO your workstation over SSH.
Lets devices push files to your workstation over SSH (TCP 22):
Local AI - private and free (Ollama) local-only
Need it if: you want AI with no cloud key and no cost - the private, local option.
How it works: Ollama is a small, free program that runs AI models locally - your own private AI on your machine, no key and no cost.
1. Install Ollama once (ollama.com). 2. Download a model or two into it (commands below). 3. In SubnetSlinger, set AI = Ollama and pick which of your models to use. The app simply talks to Ollama on your machine - nothing leaves it.
Step 1 - install Ollama. The official one-line installer from ollama.com (or grab the package there):
Step 2 - pull a model. It must be tool-capable - the Deputy drives tools, so the model needs function calling (not all do). Bigger models answer better but need more RAM/GPU (see specs). Reliable picks, start with the first:
See what you've pulled (or browse the catalog at ollama.com/library):
- In the app
- Deputy > Settings > AI = Ollama > hit refresh to list the models you've pulled, then pick one
- Needs
- 32 GB+ RAM and a GPU for a good experience (see specs above)
- Share it
- Point the Deputy at one Ollama server on your LAN and every workstation stays at minimum spec
Connect the AI to your own tool servers (MCP) Node / uv
Need it if: you'll connect the Deputy AI to your own external tool servers (MCP).
Install the toolchain your MCP servers need:
MCP servers run as local child processes over stdio - no inbound ports.
Lab throughput testing (iperf3) iperf3
Need it if: you'll measure throughput/bandwidth between two machines (Lab Testing).
Needed on both ends of a Lab Testing throughput run:
Third-party servicesAI providers & sign-in - all opt-in
Nothing here runs until you turn it on. Each row shows what it does, what it sees, and what it needs.
ollama pull, 4-10 GB for a typical one), and it must support tool/function calling.How cloud AI works: the model runs on the provider's servers (Anthropic, OpenAI, or Google) - not your machine. You paste your own API key once; then when you ask the Deputy a question, it sends your prompt plus the relevant tool output - scrubbed of secrets by the local redactor first - to that provider and shows you the answer. Nothing to download, and the model list stays current on its own. You pay the provider directly for what you use (usually pennies); we take no cut, and never see your key or your data.
The three cloud providers below work the same way - they differ only in where you get the key and the endpoint the app calls:
Anthropic (Claude) opt-in
- Does
- Deputy AI provider - answers questions about tool output, can call safe read-only tools
- Sees
- Your prompts + tool output, after the local redactor strips secrets and tokens
- Where
- api.anthropic.com
- Needs
- Your own key from console.anthropic.com (stored locally)
OpenAI-compatible endpoint opt-in
- Does
- Deputy AI provider - any endpoint that speaks the OpenAI Chat Completions API
- Where
- Whatever URL you paste - Azure OpenAI, Databricks, self-hosted vLLM/LM Studio, Groq, OpenRouter
- Needs
- Endpoint URL + API key from your provider
Google Gemini opt-in
- Does
- Deputy AI provider
- Where
- generativelanguage.googleapis.com
- Needs
- Your own key from ai.google.dev
Ollama (local) local-only
- Does
- Deputy AI provider running entirely on your machine - no key, no cloud round-trip
- Where
- localhost:11434 (or a server you point it at)
- Needs
- Ollama installed (see the add-on above)
Account sign-in opt-in
- Does
- Unlocks your paid license - sends a six-digit code to your email
- Sees
- Your email address; verifies the code you type back
- Runs when
- You tap Sign in on the Account screen
License purchase (Stripe) opt-in
- Does
- Payment for the Pro license
- Sees
- Your name + card details - Stripe handles the card, we never see it
- Where
- checkout.stripe.com, in your default browser (not embedded)
First time with AI?How to get a cloud API key
You only need a key if you want cloud AI. The free tier and local Ollama need none. Getting a key is the same idea for every provider:
| Provider | Where to create a key |
|---|---|
| Anthropic (Claude) | console.anthropic.com > API Keys |
| Google Gemini | ai.google.dev > Get API key |
| OpenAI-compatible | your provider's console (platform.openai.com, Azure OpenAI, Databricks, Groq, OpenRouter, ...) |
What does cloud AI cost?
You're billed per token by the provider - input (everything sent) plus output (the reply) - with no markup from us. Example list rates for the default Claude models (per 1,000,000 tokens; rates change, so the app pulls the current ones from subnetslinger.com/models.json). OpenAI and Gemini have their own rates.
| Model | Input | Output | Good for |
|---|---|---|---|
| Haiku 4.5 (default) | ~$1 | ~$5 | quick lookups, most questions |
| Sonnet 4.6 | ~$3 | ~$15 | multi-step investigations |
| Opus 4.8 | ~$5 | ~$25 | the hardest problems |
A typical short question is a few thousand tokens - fractions of a cent on Haiku (the default). The app shows the running cost live as it works, and local Ollama is always free.
Supply-chain integrityVerify this download is genuine
Every Linux release is signed with an offline GPG key - a full compromise of our website still could not forge a build. The files that ship next to the tarball: the detached signature, SHA-256/512 checksums, the public key, and a CycloneDX Software Bill of Materials (SBOM) - all covered by the signature.
Expect Good signature, the fingerprint below, and OK on the checksums.
- Offline Ed25519 signing key - the private key never touches an internet-connected machine
- SBOM (CycloneDX,
*.cdx.json) ships in the download set and is scanned for dependency CVEs - No telemetry. No analytics. No third-party tracking SDKs.
Cross-check that fingerprint against subnetslinger.com/download - a tampered page cannot match both.
ReferenceNetwork ports - the built-in test servers
These listen only when you use the Local Servers tool, on the standard service ports so real gear can point at your workstation. On Linux, a refused low-port (below 1024) bind falls back to a high port automatically - no grant needed.
| Service | Proto / Port | Service | Proto / Port |
|---|---|---|---|
| Syslog | UDP 514 | RADIUS | UDP 1812-1813 |
| SNMP trap | UDP 162 | TACACS+ | TCP 49 |
| TFTP | UDP 69 | FTP | TCP 21 |
| DNS | UDP 53 | HTTP | TCP 80 |
| DHCP | UDP 67 | SFTP/SCP | TCP 22 |
Trial & licensingFree, trial, and Pro
Free tier - 9 tools, forever, no account, no internet required.
Free trial - 30 days of full Pro, no card. One trial per email. Start it in-app: Account > Start trial.
Pro - $499 USD, one-time, via Stripe. Unlocks every desktop OS (Windows, macOS, Linux) for v1.x. No subscription; every dot-release inside v1 is free.
Mobile - free forever on iPhone, iPad, and Android. Companion apps, no gate.
ReferenceWhere your data lives
| Path | Contents |
|---|---|
| ~/.config/SubnetSlinger | settings & state |
| ~/Documents/SubnetSlinger | server folders (tftp/http/ftp/dns/radius/tacacs/scp), exports, captures, logs |