Lease an agent. Run one command. It builds the stack inside your own environment: Kubernetes, Ceph, Postgres HA, GitOps, observability. You keep your hardware, your cloud and your data. We run the control plane.
6 free agents · MIT licensed · revoke any lease, any time
HA Postgres. A Ceph cluster. A GitOps pipeline. Observability that actually works. The people who can do it well are scarce, and the runbooks rot the day after they are written.
Pick an agent and accept the terms. Free agents lease instantly; Team agents check out through Stripe.
→One install command deploys the agent into your environment and pairs it to your lease.
→Scheduling, updates, and policy flow down automatically. You watch it all from one dashboard.
→Add data planes as you grow, or pull a lease instantly — your data never has to move first.
Kubernetes, GitOps with Argo CD, pipelines with Tekton and Jenkins. A cluster you can hand to a team.
Prometheus and Grafana on a cluster that already exists, with a rollout check so you know it is really up.
Certificate automation with cert-manager and a Vault PKI, built for private networks where a public CA is out of reach.
MinIO, Ceph and NFS, with benchmark agents that show what your storage really delivers.
Postgres HA with Patroni and etcd, and Trino for queries over your own data.
The infrastructure bench you cannot afford to hire. Start with six free agents, no card.
Agents built for real infrastructure workloads, grouped by the team that reaches for them. The six benchmark and testing agents are free and open source (MIT); Team and Fleet agents are leased, and you get the code through your lease.
Bring your own racked, cabled, networked hardware — these agents turn it into a running cluster.
Bootstraps a Ceph cluster with cephadm — bootstrap, add hosts, apply OSDs, create pools and services, then checks cluster health. Pair with the storage benchmark agent once it's up.
Deploys an OpenStack overcloud with TripleO/director — undercloud config, instackenv, flavors/profiles, overcloud deploy, then checks overcloud status. Assumes hardware and networking are already configured.
Bootstraps a self-managed Kubernetes cluster — installs containerd (or Docker + cri-dockerd), runs kubeadm init, applies Calico or Flannel, joins workers, then verifies nodes. Hand the running cluster to the Prometheus+Grafana or cert-manager+Vault agents below.
Deploys OpenShift Container Platform / OKD via the modern Agent-based Installer — generates install-config/agent-config, builds the boot ISO, waits for bootstrap and install, then verifies the cluster. No separate Ironic provisioning host needed.
Run against any cluster above — or any cluster you already have.
Deploys MinIO object storage — standalone single-node or distributed erasure-coded cluster — generating the systemd service, installing binaries, bootstrapping buckets, and verifying cluster health.
Deploys an HA load-balancer front-end — real haproxy.cfg for L4/L7 balancing, keepalived.conf for a floating VIP over VRRP, then checks both daemons and the VIP are up. Points at any backend pool: a K8s API server, an OpenStack endpoint, a Ceph RGW cluster, a database cluster.
Day-2 operations agents — meant to run against a cluster that already exists.
Deploys the Prometheus + Grafana monitoring stack onto an existing Kubernetes cluster — real values.yaml for the official kube-prometheus-stack Helm chart, a real helm upgrade --install, Grafana datasource/dashboard provisioning, and rollout verification.
Deploys TLS certificate lifecycle automation for a Kubernetes cluster — installs cert-manager via Helm, configures a HashiCorp Vault PKI secrets engine as the CA, wires the Vault Issuer, and issues a test certificate end-to-end. Fits on-prem/private-cloud setups where a public ACME CA is unreachable.
Deploys a highly-available PostgreSQL cluster — etcd quorum, Patroni config per node, one primary with N streaming-replication replicas and automatic failover, then checks cluster status.
Deploys a distributed SQL query engine — PrestoDB or Trino, your choice — on Kubernetes via Helm, wiring up catalogs and a Hive Metastore so it can query data in place (S3, HDFS, and friends). Once it's up, hand it to the PrestoDB EKS scaling agent below, or to the database benchmark agent to measure it.
Goal-driven agent that autonomously scales a PrestoDB cluster on AWS EKS from 0 to 80 pods, monitoring readiness and coordinator registration along the way. Bring your own AWS IAM role. Pairs with the Deploy PrestoDB Trino agent above.
CI/CD tooling, deployed onto infrastructure you already have — not the infrastructure itself.
Installs a Jenkins controller, bootstraps the initial admin setup via JCasC, installs core plugins, joins an SSH build agent node, and seeds a pipeline job — then checks controller and agent health.
Installs and registers self-hosted GitLab Runners against your existing GitLab instance (cloud or self-managed) — install, Docker executor prereqs, registration (current token or legacy flow), generated config.toml — then checks runner status.
Installs Tekton Pipelines, Triggers and Dashboard onto an existing Kubernetes cluster, applies a sample Task/Pipeline, runs a test PipelineRun, then checks pod and run status. Pairs with the Argo CD agent below for the CD half.
Installs Argo CD onto an existing Kubernetes cluster, retrieves the initial admin password (never decoded or echoed by the agent), generates an Application manifest with optional automated sync/prune/self-heal, then checks sync and health status.
Describe a workload in plain English — it generates and runs fio, vdbench, iozone, and dd workloads, parses each tool's own IOPS/throughput/latency output, and produces a full side-by-side comparison report.
Describe a workload in plain English — it drives a full sysbench OLTP run against MySQL or PostgreSQL and parses TPS/QPS/latency, plus generates ready-to-run HammerDB TPROC-C scripts.
Describe a target in plain English — it drives iperf3 and netperf for raw TCP throughput, mtr for hop-by-hop path latency and loss, and wrk for HTTP-layer load, reporting each grouped by what it actually measures.
Two complete implementations — LangGraph and CrewAI — targeting REST APIs from natural language prompts, with retry logic and a side-by-side framework comparison.
Runs NFS Connectathon suites — basic, general, special, lock — across NFSv3, NFSv4, and NFSv4.1, then flags regressions against a baseline.
Runs Tempest test groups — compute, network, storage, identity, image — across two environment configs and surfaces exact failing test names and regressions.
k6-driven load and stress testing for REST and web endpoints by default, with JMeter available for JDBC, JMS, LDAP, and other non-HTTP protocols k6 doesn't cover.
We deployed our agents onto a live 3-node Kubernetes cluster and checked each result with kubectl, not the agent's own word.
Monitoring stack up and healthy.
Certificate automation installed and issuing.
Distributed SQL engine deployed and ready.
GitOps controller synced and healthy.
Leasing an agent splits responsibility cleanly: we operate the brain, you keep the body. Nothing that touches your data ever leaves your infrastructure.
The control plane issues the agent's config, rotates credentials, and holds the dashboard — but never touches your workload data directly.
The leased agent runs inside your VPC, cluster, or bare metal. Compute, storage, and egress stay on infrastructure you already control.
Every plan includes the hosted control plane. You bring the compute for your data plane. Fleet isn't "the agents we gate" — it's for running agents as a fleet: many clusters, many regions, many subscriber orgs, one tenancy.
Benchmarking & testing agents only. Good for evaluating the lease model with zero risk to any cluster.
Every deployment, HA, observability & security agent that runs on infrastructure you already have — minus the bare-metal platform rollouts. Built for standing up and running real infrastructure on environments already racked and networked.
For running at fleet scale — many clusters, many regions, many subscriber orgs under one tenancy — plus the deploy agents that stand up bare-metal platforms from scratch.
Free agents are open source (MIT). Team and Fleet agents are provided under the Agent4Lease Customer License while your lease is active. View licenses
The same lease model, pointed at the jobs your team dreads. Tell us which one you want first.
Agents are plain Python that runs on a Linux host or Kubernetes cluster you control: your own racks, your cloud account, or a GPU neocloud. You keep your own account and billing with the provider. Agent4Lease doesn't resell compute.
Prometheus + Grafana, cert-manager, Trino and Argo CD passed on a live 3-node GKE cluster, each checked with kubectl. More Kubernetes agents are verified one at a time, and results are published as each completes.
CoreWeave, Lambda, Crusoe, Nebius, RunPod and similar. Benchmark and deploy agents aimed at GPU clusters, running in your own account with the provider. A provider is listed as verified only after the agents have run there.
Your own hardware, from racked servers to full platform builds. Fleet agents cover bare-metal rollouts, and we scope the work with you.
Provider names are trademarks of their owners and are used only to describe where agents can run. No affiliation with or endorsement by any provider is implied. Interested in a provider that isn't listed? Tell us.
Pick an agent and we'll generate a one-line install command tied to your lease. Run it on the machine that will host the agent.
Start here before reaching out — most questions are answered by the walkthrough or your own lease status.
A step-by-step slide deck: lease an agent, save your subscriber ID and pairing token, run the install command, and check status — plus a full worked example.
View the tutorialPaste the subscriber ID you were given at lease time to see every agent you've leased and whether its data plane is online.
Open the dashboardIf the tutorial and dashboard don't answer it, send us a message — it comes straight to our inbox.
Early-stage — the control plane is live and the agent catalog is validated. Here's the shape of the bet.
Subscribers pay to keep an agent leased and scheduled, not to own a script. Revenue compounds with fleet size, not headcount.
We host the brain; subscriber data never leaves their infrastructure. That's the thing a generic script or open-source agent can't offer on its own — and it's what makes this sellable to security-conscious infra teams.
The lease API runs on Cloud Run + Cloud SQL today, with Stripe billing live for the Team plan. The six benchmark and testing agents are free and open source. Per-agent test results on Kubernetes are published as each one completes.
Early-stage conversations welcome. If the control-plane/data-plane model is interesting to you, let's talk.
Get in touch