AI-powered operations for Linux · Development ongoing · Available to try soon

Your AI sysadmin.
Always on watch.

AISysAdmin monitors system health, investigates faults, spots security risks, and turns noisy server data into clear action before small issues become outages.

Meet your AI sysadmin →See what it monitors

12system probes

4alert channels

1operational view

web-01 / latest report complete

System analysis

2 issues need attention

14:32 UTC

CRITICALstorage

Filesystem /var is 96% full

Free space is below the safe operating margin.

/dev/sda2 96G 92G 4G 96% /var

HIGHsecurity

Repeated SSH authentication failures

53 failed attempts from three sources in the last hour.

Human review requireddu -x -h --max-depth=1 /var | sort -h

Suggested for investigation. Never executed by the agent.

✓ shell=False

✓ secrets redacted

01 / OPERATING MODEL

From raw signals to
clear next steps.

A tireless operational loop that helps teams understand problems sooner, respond faster, and keep services dependable.

Monitor

Check the system signals that reveal emerging faults, pressure, and unusual activity.

Investigate

Connect diagnostic evidence and logs to explain what is happening and why it matters.

Protect

Surface security risks and suspicious behavior before they disappear into operational noise.

Guide

Deliver prioritized findings, practical next steps, and timely alerts to the people who can act.

02 / TRUST BY CONSTRUCTION

Powerful insight.
Predictable boundaries.

AISysAdmin is built to deliver useful operational intelligence without becoming another source of risk. Diagnostic access is deliberately constrained, verified, and run by a dedicated unprivileged user.

✓ Exact command and argument allowlist
✓ Clean environment, closed stdin, fixed working directory
✓ Output limits, timeouts, and cross-process run lock
✓ Suggestions displayed for humans—never piped back to execution

EXECUTION_POLICYENFORCED

01probe = registry.get(request.name)02assert request == probe03assert executable in TRUSTED_PATH04subprocess.Popen(05probe.argv, shell=False,06stdin=DEVNULL, env=SAFE_ENV07)

Unknown command deniedFail closed · no fallback execution path

03 / CAPABILITIES

One agent.
A wider field of view.

Health, security, investigation, and alerting brought together so your team can stay ahead of operational trouble.

Fault detection

Spot full filesystems, memory pressure, failed services, and network interfaces that have dropped.

Evidence-led analysis

Turn journal events and diagnostic output into concise summaries with the source evidence attached.

Security signals

Surface repeated SSH failures, unexpected listeners, suspicious login activity, and likely intrusion attempts.

Alerts that fit your stack

Notify operators by email, webhook, syslog, or SNMP when a finding crosses your severity threshold.

Context-aware memory

Reuse fresh observations only when host context, kernel, boot ID, instructions, and probe set still match.

Replaceable intelligence

Use the OpenAI Responses API, the offline rules engine, or your own AnalysisProvider subclass.

04 / WEB ADMIN

See the system.
Shape the agent.

Configure server context, provider, probe groups, cache policy, and alert thresholds from an authenticated, local-first FastAPI console.

LOCAL FIRST Loopback-only by default. Remote access belongs behind an SSH tunnel or authenticated HTTPS proxy.

GET STARTED

Put an AI sysadmin
on watch.

Give your team earlier warnings, clearer answers, and practical guidance for healthier, safer systems.

INSTALLPython 3.11+

$ pip install -e .$ aisysadmin list-probes$ aisysadmin check -c config.yamlReview the safety model →

Your AI sysadmin.Always on watch.