Features

Everything your infrastructure
needs to run itself.

Seven systems that work together. Each one solves a real problem. Together, they change how IT operates.

01

AI Chat

Talk to your infrastructure in plain English. Aorka knows what's running on every endpoint, what changed recently, and what's been tried before. When action is needed, it writes the script, explains it, and waits for you.

  • — Powered by Claude — context-aware, not just autocomplete
  • — Full environment context injected into every conversation
  • — Microsoft 365 integration — query mailboxes, licenses, Entra ID
  • — Conversation history preserved — pick up where you left off
you: Which servers have less than 10GB free on C:?
aorka: Checking known disk facts across your endpoints...
searched 47 endpoints — 3 matches
aorka: Three servers are low:
  FS-01: 4.2 GB free (last checked 2h ago)
  DC-02: 7.8 GB free (last checked 6h ago)
  APP-03: 9.1 GB free (last checked 1h ago)

FS-01 is critical. Want me to check what's consuming space?

FS-01 — Windows Server 2022 Standard

Confidence: 98% · Verified 2 hours ago

Print Spooler service: Running

Confidence: 95% · Verified 4 hours ago

Ricoh MP C3004 driver v2.4.1

Confidence: 71% · Verified 6 days ago · decaying

Backup agent: Veeam 12.0.0.1420

Confidence: 34% · Verified 45 days ago · stale

02

Living Knowledge Base

Facts are collected directly from endpoints — not typed into a wiki. Each fact has a confidence score that decays over time. Stale data surfaces itself. Your documentation is never out of date because it's never written by hand.

  • — Automatic collection from Windows and network endpoints
  • — Confidence scores with statistical variance tracking
  • — Semantic search via vector embeddings — find by meaning, not keywords
  • — Hierarchical unit tree — organize by client, site, or function
03

Script Safety Pipeline

Three layers between the AI and your machines. A regex command filter classifies every operation. An AI evaluator scores blast radius. An approval system gates execution by risk tier. Nothing runs unchecked.

  • — Regex command filter: hard classification of every cmdlet and operation
  • — AI risk scoring: static analysis + LLM evaluation of intent and impact
  • — Tiered approvals: read-only auto-runs, writes need approval, dangerous blocked
  • — Unit read-only locks — new clients start locked, unlock when you're ready
LAYER 1
Regex Command Filter

Pattern-matches every cmdlet: Get-* → read-only, Restart-* → write, Format-Volume → blocked

LAYER 2
AI Risk Evaluation

Claude analyzes intent, scope, and reversibility. Scores 0–100.

LAYER 3
Approval Gate
Read-only → Auto Write → Approve Dangerous → Blocked

Restart-PrintSpooler

Clears stuck queue, restarts service, verifies recovery

94%

847 runs

Get-DiskSpaceReport

All volumes, free %, largest files, trend

99%

2,341 runs

Repair-ADReplication

Force sync, check status, report lag

87%

412 runs

Clear-TeamsCache

Stop Teams, purge cache, restart

72%

1,156 runs

04

Proven Script Library

Stop generating untested scripts. Aorka's library has thousands of parameterized scripts — every one security-evaluated and scored by real execution history across real infrastructure. When a script fails, its score drops. When it succeeds, it rises. You're not the guinea pig.

  • — Thousands of scripts, scored by real success and failure data
  • — Parameterized for general use — works across any environment
  • — Growing constantly — every new solution becomes available to everyone
  • — Semantic search — describe the problem, find the script
Domain Admin — contoso.local
Last accessed 3d ago
FortiGate Admin — fw-01.mgmt
Last accessed 12d ago
SQL SA — sql-prod-01
Last accessed 31d ago
MFA verification required to view secrets
05

Credential Vault

Stop putting passwords in shared spreadsheets. Credentials are AES-256 encrypted at rest, MFA-protected on access, and injected into agent sessions at runtime. They never touch disk on the endpoint.

  • — AES-256-GCM encryption with per-credential keys
  • — MFA required to view or use secrets
  • — Break-glass access with full audit logging
  • — Webhook alerts on credential access
06

Bulk Deployment

Roll out agents to an entire client environment in minutes. Deploy through your existing tools — GPO, Intune, NinjaRMM, JumpCloud, or anything else that can push an MSI. Or push directly from the console. Watch them come online in real time.

  • — MSI installer — deploy via GPO, Intune, NinjaRMM, JumpCloud, etc. — anything that deploys MSIs
  • — Active Directory and Entra ID sync — auto-discover machines
  • — Remote push from the console — no RDP, no manual steps
  • — Outbound-only connections — zero firewall changes at client sites
DC-01
online
FS-01
online
APP-03
installing...
WS-04
pending
WS-05
pending
Source: Active Directory 2 of 5 deployed
Phase 1: Pre-checks completed
disk space ✓ services ✓ backup ✓
AI Gate: "All pre-checks passed. Safe to proceed."
Phase 2: Patch deployment running
DC-01 ✓ FS-01 ◌ APP-03 —
Phase 3: Validation waiting
07

Orchestration

Multi-phase runbooks that execute themselves. Define the phases, set the gates, and let the engine handle sequencing, fan-out, and failure propagation. Runbooks that actually run.

  • — DAG-based execution — phases, steps, dependencies
  • — AI gates — Claude evaluates results before proceeding
  • — Fan-out across multiple endpoints in parallel
  • — Automatic failure propagation — one failure halts downstream work
Bonus

MCP Server

Already using Claude Code, Codex CLI, or Gemini CLI? Connect them directly to Aorka via the Model Context Protocol. Same safety pipeline. Same approval flows. Your existing AI tools, now with infrastructure awareness.

# .claude.json
{
"mcpServers": {
"aorka": {
"url": "https://aorka.com/api/mcp",
"headers": { "Authorization": "Bearer ..." }
}
}
}

Ready to see it?

We'll walk you through the platform with your own infrastructure. Thirty minutes. No slides.

Request a demo