Amigo CLI by zapit
Your best friend — ask in plain English and it investigates your devices, explains the cause, and recommends the fix.

An AI SSH & API terminal for the future

Let Amigo CLI Troubleshoot·Report·Configure

Buy a license — $19.99 Choose a plan — from $19/mo

The terminal itself is free — download it here.

SSH terminal

SSH — read-only by default; the assistant finds the cause and recommends the fix, you run it. Switch to write mode and it configures for you too.

API devices

API devices — Meraki, Firepower (FMC), ISE, Catalyst 9800; same assistant, every change gated behind your approval.

what you get

Built for network engineers

An SSH terminal that does the digging for you — safely, and on your own AI key.

AI that investigates

Ask in plain English. It runs the right read-only show commands, reads the output, and tells you the cause.

Read-only by design

The AI recommends changes; you approve and run them. It never changes a device on its own.

Bring your own AI key

Use your own provider key (Anthropic, OpenAI…). Your data goes to your account — no AI fee on top.

Import from SecureCRT

Bring your saved sessions and folder tree across in one click.

Automatic session logging

Every session saved to a plain-text file — grab a show run for TAC without digging.

Secrets stay put

Passwords and keys live in your OS keychain, and are designed to be stripped before anything reaches the AI.

download

Get the app for your platform

The app is free to download and use as a full SSH / API terminal — no key needed. The AI assistant unlocks with a one-time license (below).

Windows

Windows 10 / 11

64-bit · Microsoft Store

Get it from Microsoft Store
macOS

macOS

Apple silicon & Intel · .dmg

Coming soon
Linux

Linux

AppImage · .deb

Coming soon

Installed and kept up to date through the Microsoft Store, so you always have the current version. macOS & Linux builds are coming soon.

pricing

Unlock the AI assistant

The terminal is free, forever. The AI assistant — the part that investigates your devices, explains the cause, and writes the fix — unlocks with a one-time US$19.99 license — a one-off purchase for one device, with nothing recurring. Bring your own AI account, or run a model on your own hardware, and that is all you ever pay. If you would rather not manage an AI account, Amigo AI plans are an optional monthly subscription — cancel any time.

Buy a license — $19.99

Your license key arrives by email right after checkout — paste it into Settings → License in the app. Secure checkout by Polar.

Next: choose how the AI is powered ↓

then choose how the ai is powered

Your own key, or ours

The license turns the assistant on. It still needs a model behind it — and you pick which. Both work the same inside the app.

Bring your own key

No extra cost from us. Paste an Anthropic, OpenAI, Azure or Google key into Settings. You're billed by them, at their prices, on your own account.

  • Nothing more to pay Zapit
  • Your usage stays on your own provider account
  • Also works with a model on your own machine — Ollama, LM Studio

Amigo AI — no key needed

From $19/month. We handle the AI account. Nothing to sign up for, nothing to configure — pick Amigo AI in Settings and ask.

  • No AI provider account, ever
  • Your license key is all the app needs
  • Change plan or cancel any time

Amigo AI plans

Priced at roughly what the AI costs us to run. Every plan uses the same assistant — the difference is how much you can ask per month.

Starter — $19/mo

For one engineer, some of the time. Comfortable for day-to-day questions and the occasional investigation.

Pro — $49/mo

For an engineer who lives in it. Deeper investigations, more device output, more often.

Max — $99/mo

For heavy daily use, or when you'd rather not think about it at all.

Choose a plan

You'll pick Starter, Pro or Max at checkout. A plan needs the $19.99 license too — the license turns the assistant on, the plan powers it.

Buying for a team?

Buy a license and a plan for each engineer. They each get their own key, their own allowance, and one person running hot never slows anyone else down. Same checkout, one per seat.

What happens if you run out?

The app warns you before you get there, and shows what's left in Settings → AI. If a month does run out, the terminal, your saved connections and your own-key option all keep working exactly as before — and you can switch to your own provider key at any moment.

set it up

Get running in two steps

Quick start.

1 — Download and install amigo.   2 — Open Settings → AI and add your own provider key (Anthropic, OpenAI…). That's it — it works on your device and offline.

Common questions

What are the system requirements?

Windows 10/11 (64-bit), macOS on Apple silicon or Intel, or a modern 64-bit Linux. Amigo CLI needs the internet for the AI assistant — the terminal itself runs offline.

How do I connect my own AI? (bring your own key)

You plug in your own provider key, so your data goes to your account. Create an API key (e.g. Anthropic or OpenAI), open Settings → AI, pick your provider, paste the key, and Save. The key is hidden as you type and stored in your OS keychain — not in a file or a log, and not sent to us.

Do I need to be online to use it?

Only for the AI assistant. The terminal itself works offline — a network outage won't lock you out mid-work.

Are my device data or passwords sent to the AI?

Passwords, keys, and enable secrets are designed to be stripped before anything is sent, and credentials are kept out of the AI path by design. Everything sensitive stays in your OS keychain on your machine.

Is Amigo CLI an alternative to PuTTY, mPuTTY, or SecureCRT?

Yes — Amigo CLI is a free tabbed SSH terminal like PuTTY or mPuTTY, with a saved-connection tree like SecureCRT (and you can import your SecureCRT sessions in one click). On top of that it adds what none of them have: an AI assistant that investigates your devices, explains the cause, and recommends the fix.

What devices and platforms does it work with?

Anything you can SSH into — Cisco IOS / IOS-XE / NX-OS, Juniper, Arista, Linux servers, and more. It also connects to API-based platforms: Cisco Meraki, Firepower (FMC), ISE, and Catalyst 9800.

Questions? Email [email protected].

releases

What's new

v1.1.6 — August 2026

Network tools, Microsoft sign-in, and your own AI model.

New

  • Map your network — scan a site and see what is there and how it connects.
  • Trace a path — follow the route hop by hop, with a plain verdict on whether the loss is real.
  • Ping sweep and port scan — is this address in use, and what is it listening on?
  • Sign in with Microsoft for Azure OpenAI — no key to paste (guide below).
  • Use your own AI model, on your machine or your own server (guide below).

Fixed

  • Clicking a device on the map now works with an ordinary mouse.
  • Opening a device that is already open jumps straight to it instead of asking.
  • The AI settings screen no longer says a feature is missing when it is there.
v1.1.5 — August 2026

Reading long output from real devices, reliably.

Fixed

  • A device that pauses at --More-- no longer swallows the answer, and Amigo says plainly when output was cut short.
  • Mangled output is now noticed: a read is retried once, a change never is — that always goes to a human.
  • | include A|B|C is no longer mistaken for a change.
  • About 15 MB smaller — binaries for other platforms removed.
v1.1.0 – v1.1.3 — earlier releases

First public release — the AI network terminal, read-only edition.

New

  • AI assistant that investigates your devices and explains issues in plain English.
  • Approve-every-change safety model — the AI doesn't change a device on its own.
  • Designed to strip secrets before anything reaches the AI; keys stored in your OS keychain.
  • SecureCRT session import, automatic session logging, tabbed SSH, keyword highlighting, encrypted vault.
step-by-step

Use Azure OpenAI with your Microsoft account

Sign in with your work account instead of pasting an API key. Steps 1–4 are done once, in the Azure portal, by whoever administers your Azure subscription. Step 5 is in Amigo.

In the Azure portal (one-time setup)

  • 1. Have an Azure OpenAI resource. In portal.azure.com, create or open an Azure OpenAI resource (also shown as Azure AI Foundry). Note its endpoint URL from Keys and Endpoint — you will paste it into Amigo.
  • 2. Deploy a model. Open the resource, go to Model deployments and deploy a chat model. Note the deployment name you give it — Amigo asks for that name, not the model's name.
  • 3. Give the person permission. This is the step that is usually missed. On the resource, open Access control (IAM) → Add role assignment, choose the role Cognitive Services OpenAI User, and assign it to the user (or a group they are in). Without this role the sign-in works and every question then fails.
  • 4. If your tenant requires admin approval. The first sign-in asks to “Access Cognitive Services on behalf of you”. If your organisation blocks users from consenting, an administrator approves that request once, for everyone. Amigo requires no other permission, and it never asks for access to your mail, files or directory.

In Amigo

  • 5. Open Settings → AI, set Provider to Azure OpenAI, set How you sign in to Sign in with Microsoft, paste the endpoint from step 1 and the deployment name from step 2, then press Sign in. Your browser opens, you sign in with your work account, and the tab closes itself. Settings then shows the account you are signed in as, and a Sign out button.

If it does not work

  • “You do not have access” after signing in — step 3 was not done, or was done on the wrong resource. The role must be on the Azure OpenAI resource itself.
  • The sign-in will not accept your account — a personal Microsoft account cannot hold an Azure role. Use the work or school account that owns the subscription.
  • Amigo says it blocked the address — the endpoint is not an Azure one. Check it against step 1; it ends in .openai.azure.com or .services.ai.azure.com.
step-by-step

Use your own AI model

Run the AI on your own computer, or on a server your company controls. Nothing you ask reaches us or any AI provider — it goes only to the address you enter.

On your own computer

  • 1. Install a model runner. Ollama is the simplest. Install it, then pull a model — for example ollama pull qwen2.5:7b. Leave it running; it listens on http://127.0.0.1:11434.
  • 2. Point Amigo at it. Settings → AI, set Provider to LLM, and put that address in Where it is listening.
  • 3. Press Find models. Amigo asks the server what it has and lists what it finds.
  • 4. Pick the model in the chat box, next to Send. That is it — ask a question as normal.

On a server your company runs

  • Same steps, with two additions. The address must be https, and you must confirm the server once: enter the address, press Confirm this address as my server, and it joins Permitted servers. Amigo will not send anything to an address you have not confirmed — that is deliberate, so a typo cannot send your session somewhere unintended.
  • If the server needs a key, paste it in Key for your server. It is stored in your operating system's credential store, the same as every other key.
  • If the server uses a self-signed certificate, Amigo stops and shows you its fingerprint before trusting it. Check the fingerprint before you accept.

Worth knowing before you rely on it

  • Small models are much less reliable. They can state things that are not true, and may answer without reading your equipment at all. Every answer Amigo gives shows what it actually ran — if there is no “what I did” underneath an answer, nothing was read, and the answer is the model guessing. Check anything before you act on it.
  • A model on your own computer is slow without a graphics card — expect to wait minutes, not seconds, for an answer.
  • The safety rules do not change. Observe still runs nothing, and read-only still only reads, whichever AI you choose.