Connect your targets
Add a saved or temporary SSH host using a command that already works for you. Kubernetes clusters use a direct API workflow with a vault-encrypted kubeconfig, so they do not need an SSH server or a separate plugin on every node.
AI-powered server administration
Manage any server like a pro,
with AI by your side.
Natural-language tasks across Linux, FreeBSD, macOS, Windows and Kubernetes. Review the plan, approve, and execute securely through your system SSH.
curl -fsSL https://raw.githubusercontent.com/cryptofuture/intentaiops/main/scripts/install.sh | sh
Intent AI Ops
ACTIONS [current]
> Browse/search hosts
Run one task on multiple hosts
Install/update plugin on multiple hosts
Browse Kubernetes clusters
Multi-host run history
Add host
Temporarily add host
Add Kubernetes cluster
Set default administrator email
Change vault passphrase
Quit
HOSTS
Search:
HOST DETAILS
Tab pane Up/Down select Enter open / search Esc back q quit
Learn how Intent AI Ops works
Operations with a review step
Describe the outcome. Stay in control of every command.
Intent AI Ops is an open-source interactive CLI for administrators who want the speed of natural-language automation without handing an AI unrestricted access to their infrastructure. It runs on your administration computer, connects with the system OpenSSH tools you already use, and turns a request into a structured, inspectable plan.
Ask in plain language
Request an outcome such as installing nginx, checking host health, applying current-release updates, or deploying a supported application. Intent AI Ops gathers relevant inventory and asks your local Codex CLI to prepare the task.
Review, approve, execute
Inspect the proposed commands, warnings, dependencies, affected paths, and revert actions before anything runs. Approve the complete plan, review commands individually, or cancel it. Conservative approval defaults keep the operator in charge.
Built for real administration
One terminal for routine work across your estate
Use the same workflow for a single machine or a group of hosts. Each target keeps an independent result, so a failed host can be diagnosed, reverted, and retried without rerunning work that already succeeded elsewhere.
Cross-platform by design
Manage Linux, FreeBSD, macOS, Windows, and Kubernetes targets. Eligibility is based on the detected operating system, distribution, architecture, service manager, virtualization, container environment, and runtime policy—not on a one-size-fits-all script.
Verified common tasks
Choose cataloged routes for host diagnostics, maintenance, runtime setup, static sites, and popular applications including WordPress, Drupal, Joomla, Moodle, n8n, Ghost, and more. Unsupported choices stay hidden when a target is not eligible.
History, retries, and reverts
Per-host SQLite history records requests, plans, progress, command output, errors, and rollback steps. Retry with earlier failure evidence and optional correction instructions, or run the saved revert plan without overwriting the original record.
System SSH, not a hidden tunnel
Connections use your local ssh and scp binaries and their familiar configuration. When automation is not the right tool, open a real interactive SSH shell and continue the investigation manually.
Security boundaries that make sense
Your credentials are not part of the prompt.
Intent AI Ops keeps SSH details and saved credentials in a strong local vault. Saving an SSH password or key passphrase is optional and off by default, while one-time sudo passwords are never stored. Sensitive values are removed from inventory, diagnostics, logs, and task history before context reaches Codex.
The AI does not receive your vault, SSH credentials, private keys, tokens, connection strings, or generated application passwords. Codex also runs locally through your authenticated CLI, so Intent AI Ops itself does not require a separate OpenAI API key.
- No per-host AI agent. Remote execution uses the signed Intent AI Ops Netdata plugin.
- Explicit approval. Commands are displayed before execution and can be cancelled.
- Auditable outcomes. Results and recovery information remain attached to each host.
- Apache 2.0 licensed. Inspect the implementation and contribute on GitHub.
Get started
Bring an SSH connection and your signed-in Codex CLI.
Install Intent AI Ops on the computer you administer from. The bootstrap installer provides Node.js 24 when needed. On first launch, create a local vault, add a host, inspect its capabilities, and try a consultation by prefixing a question with ?—consultations never propose, approve, or execute commands.