Skip to content

AI

OpsPilot has two separate AI modes. AI Providers means you supply an API key or a sign-in and the AI lives in OpsPilot's own panel. AI Assistants means an app you already pay for — Claude Desktop, ChatGPT, VS Code Copilot Chat — connects to OpsPilot over MCP and drives it from there. Both work on SSH and Local Console tabs.

AI Providers AI Assistants
What it is You supply an API key or sign-in You connect an app you already pay for
Who drives OpsPilot's own AI panel Claude Desktop, ChatGPT, VS Code Copilot
Cost model Pay per request, or free with a local model Model usage covered by your Claude, ChatGPT or Copilot plan
Offline capable Yes, with Ollama or self-hosted No — those apps are cloud services
Approval gate Yes Yes, identical
Redaction Yes Yes, same profiles

Whichever mode a proposal arrives through, it enters the same approval queue, gets the same risk tier, meets the same Command Safety rules, and its context passes through the same redaction layer first.

Use Providers when you want the AI inside OpsPilot, or when you need a deployment that never reaches the internet. Use Assistants when you would rather work in a chat app you already have open, or drive the estate from your phone. Both at once is fine — they are independent settings, not a choice.

AI and your license

Both modes need an active free trial or subscription:

  • During the 15-day trial, AI is on for 2 of your connections at a time, and you choose which. AI Assistants and the ChatGPT tunnel work too.
  • After the trial without a subscription, AI is off in every session, and AI Assistants and the tunnel are off. The terminal keeps working, and your providers, assistants and profiles stay configured.
  • With a subscription, AI is available on every connection, and comes back by itself where you had it on.

Losing the license never cuts off an answer: an AI answer already under way finishes, nothing it proposes runs, and new requests are refused. See Free trial and limits and Subscribe.

Pages in this section

  • AI Providers — the ten built-in providers, how to enter a credential, test it and switch model per session.
  • Offline with Ollama — a local model and an offline license, so OpsPilot works with no route to the internet at all, and what to expect from a local model.
  • Self-hosted & enterprise — Azure OpenAI in your own tenancy, and any OpenAI-compatible endpoint you already run.
  • AI Assistants (MCP) — the four assistants, the six tools they get, and how the safety model applies to them.
  • Remote & mobile operation — the ChatGPT Web/Work tunnel, and what works well from a phone.
  • Using the AI panel — the day-to-day task, step by step: ask, read the proposal, then approve or dismiss.
  • What gets sent — exactly what a provider or an assistant receives, what never leaves the machine, and what licensing sends.

Private or VPN-only is not air-gapped

With a cloud provider your workstation still reaches the internet, even though your servers do not. That is private or VPN-only operation. Air-gapped use needs a local model and a license that needs no network: offline activation or a deployment license.

See also