Who it is for¶

OpsPilot is aimed at engineers who operate infrastructure interactively — people who spend their day in a terminal against machines other people depend on — and at the security and platform teams who have to be comfortable with what those engineers are running. Six groups get distinct value from it.

Site reliability engineers¶

Incident response where the first ten minutes go on reading logs across four hosts. Open a session per host, ask the AI panel what it sees, and it reads them in parallel, correlates, and proposes the check you were about to type — with secrets stripped before anything leaves the machine. The quick actions in the AI panel (Analyze Error, Explain Command, Review Config, Optimize) cover the questions you ask most often during an incident.

DevOps and platform engineers¶

Release validation, rollout verification and the long tail of "why is staging different from production". Because a Command Safety profile and a Data Handling profile attach per connection or per group, the same assistant can be permissive in the lab and strict in production without anyone remembering to change a setting first.

Network engineers¶

Telnet and serial console access to switches, routers and firewalls that will never run an agent, in the same window as SSH sessions to them, where the AI helps read configuration diffs and interface state. The connection types that matter here are the ones AI tooling normally ignores — see Connection types.

Developers¶

Deploy and test the code you just wrote against real infrastructure, without waiting for a platform team ticket, and inside the guardrails the platform team configured. OpsPilot does not try to replace the tool you write code in; it owns the half that tool cannot reach — the environment the code has to run in.

Security and compliance teams¶

A demonstrable boundary between an AI and a production shell, a redaction layer that runs on one code path before egress, and a click required on every command that counts as dangerous, with a typed justification on by default. What makes this reviewable is that the boundary is architectural rather than configurable: there is no setting, provider or prompt that hands a model a shell. What may run without a click is set per Command Safety profile, so it can be reviewed like any other policy. See Security model.

Regulated and disconnected environments¶

Banking, defence, healthcare, utilities, industrial control, and air-gapped estates. This is the audience that has been told for years that AI assistance is not available to them, because the hosts cannot reach a model and because handing a model a shell was never going to pass review. OpsPilot addresses both halves. With a local model via Ollama, and an offline license file or an organisation-wide deployment license, nothing leaves the workstation at all — see Licensing for IT.