AIPrunr
Reclaim wasted cloud and datacenter spend, based on real network traffic.
AIPrunr reads your cloud inventory, billing data and network flow logs to find what nobody is actually using, and shows the dollar value of each finding. It is agentless and read-only, runs as an appliance inside your own network, and sends nothing to us. The live demo is the real AIPrunr dashboard, running on simulated data.
Connects through read-only cloud and datacenter APIs. Nothing to install on your workloads.
Separates real business traffic from background noise such as DNS, NTP and monitoring.
Runs inside your network with no telemetry, and can run in a subnet with no internet access.
A resource is a kill candidate only with traffic and billing evidence. Anything unproven is shown as potential, never confirmed.
The cloud waste problem
Idle spend keeps billing
Forgotten test environments, orphaned disks and oversized clusters keep running, and keep billing, long after anyone uses them.
Zombie servers
Metric tools see low CPU, but miss servers that are running yet serve no business traffic.
Agent fatigue
DevOps and SecOps teams reject invasive agents. You need visibility without the installation headache.
The dirty tag problem
Tools that rely on billing tags cannot see untagged assets and call tagged-but-idle assets active. AIPrunr looks for life in the traffic instead.
How it works
Connect
Give AIPrunr read-only access to AWS, Azure, Google Cloud or OCI, or to your datacenter management (for example vCenter or Hyper-V).
Ingest
It pulls inventory, billing and cost data, and VPC, VNet or NSG flow logs.
Analyze
The engine filters background noise (DNS, NTP, monitoring) and combines 30-day metric trends with recent traffic to find true idleness.
Act
You get a kill list with dollar values and the evidence behind each item, as CSV and PDF reports. Your team decides and takes the action, or uses the one-click Kill and Unkill switch, available on request.
Where it looks
Public cloud
- AWS, Microsoft Azure, Google Cloud and Oracle Cloud (OCI), with flow-log analysis on all four.
- Compute, containers (including GKE), storage, backups, databases, networking and managed services.
- GenAI and GPU spend (for example Vertex AI) and BigQuery query costs.
- Unattached volumes, unused IP addresses and old snapshots that keep billing.
- AWS Organizations: discover every member account from the management account.
Datacenter and private cloud
- VMware vCenter, Microsoft Hyper-V, OpenStack and HPE OneView.
- Physical servers through Redfish and IPMI, plus SSH, SNMP and Xen sources.
- Runs fully inside your network, with no data egress.
- Coverage varies by source: where a category is not yet collected, the dashboard says so instead of showing zero.
Read-only by design
Least-privilege access
Discovery and cost data use read-only permissions, for example ce:GetCostAndUsage and ec2:DescribeInstances on AWS. AIPrunr never asks for write or delete rights.
No egress
Credentials, inventory and findings stay on the appliance. No telemetry, crash reports or metadata are sent to us; you can block all outbound traffic except your cloud provider APIs.
Encrypted storage
Data lives in your own PostgreSQL database; cloud credentials are encrypted with AES-256-GCM before they are stored.
Honest numbers
Costs come from billing data. A resource with no recorded cost is never a kill candidate, and savings that are not yet confirmed are labelled as potential.
On request: one-click Kill and Unkill switch
For teams that want AIPrunr to act on the kill list, not only report it. Because it needs write access that the standard read-only appliance never has, we set it up with you on request and agree its scope first.
Kill
Stops an idle workload you have approved, using just-in-time credentials that expire after the action.
Snapshot first
A snapshot is kept before anything is stopped, so the workload can come back exactly as it was.
Unkill
Restores the workload in one click if someone turns out to need it.
Your approval, every time
Nothing is stopped without a named person approving it, and the resource owner is notified first.
Running it
- Ships as a Linux x86-64 appliance. On Windows or a Mac, run it in a Linux VM, for example Ubuntu 22.04 or 24.04 on Hyper-V or VMware.
- Needs PostgreSQL 14 or later and Node.js 18 or later. If Docker is installed, the start script can create a local PostgreSQL container for you.
- A setup wizard protected by a one-time token guides the first configuration; a read-only preflight check tests resources, ports and database access first.
- Starting size: 4 vCPU, 16 GB RAM and 100 GB SSD for an evaluation; 8 vCPU, 32 GB RAM and 250 GB for multiple accounts.
- Licences are bound to your billing accounts and named datacenter systems.
Licensing
14-day proof of concept
$0. Read-only access to your own cloud or datacenter, unlimited discovered assets, and PDF and CSV report export. No credit card required.
Sovereign Enterprise
Annual licence for continuous waste detection across your cloud and on-premise estate, with priority support. The one-click Kill and Unkill switch is available on request. Deploys inside your own network.
MSP & sovereign cloud
Partner and custom pricing for managed service providers and large groups, including air-gapped and on-premise appliance options.
Try AIPrunr with your own data
See the sample demo first, then tell us about your environment. We agree the scope with you before anything is installed.