About NetStacks
A desktop terminal for people who run networks, made by people who run networks, with the AI kept on a short leash.
Where it came from
The tools most network engineers use every day were designed before the network had an API, before topology lived in NetBox, and long before anyone expected a terminal to understand what a BGP neighbour table meant. The terminal stayed a terminal. Everything else moved into browser tabs, spreadsheets and other people's heads.
NetStacks started as a replacement for one of those terminals and grew into the workspace around it: sessions and credentials, topology, change control, documents, and an assistant that is given the same context the engineer has.
What it is for
Connecting to devices over SSH, Telnet and SFTP without a scattered library of session files. Reading output faster, with an assistant that can see it and suggest the next command. Making changes as a method of procedure with checks that can say no. Writing down what happened, on the device, where the next person will find it.
The free desktop application does all of that on one machine. The NetStacks Controller adds what a team needs: server-side credentials, shared device memory, audit logs and single sign-on.
The rules
Your model, your keys. There is no NetStacks AI account and no relay. The application talks to the provider you configure, with your key, and the request it sends is something you can read.
Your data stays where you put it. Sessions, credentials, documents and history live in an encrypted local database, or on your own Controller. The security page lists exactly what the application sends anywhere, and to whom.
Configurable, with exceptions on purpose. Shortcuts, prompts, presets, themes and automation are all yours to change. The few things that are fixed, such as the agent operating guide and the dangerous-command list, are fixed to keep an autonomous assistant honest, and the settings page says so.
The terminal comes first. Every feature is judged on whether it makes the terminal better or gets in its way. If it cannot be reached from the keyboard, it is not finished.
How to reach us
Bugs and feature requests: support. Security reports: see the security page for the disclosure process. Teams evaluating the Controller: contact us.
How we work
Release often
Small, frequent releases with a public changelog. A fix that is ready ships in the next build, not the next quarter.
Show the real thing
Every screenshot and recording on this site is the shipping application against lab devices. No mock-ups, no staged interfaces drawn in a design tool.
Do not break the import
Moving from SecureCRT should take an afternoon. Existing session libraries import as they are, folders included.
Say what the AI is doing
Every prompt is visible, every automatic AI feature has a switch, and anything that runs a command on a device asks first.