Agent4Lease is an attempt to fix that with a control-plane / data-plane split:
- Authors publish an agent (repo + a lightweight review) - Runners lease it directly into their own Kubernetes / OpenStack / bare-metal environment — the workload runs on their infra, not ours - The control plane only handles scheduling, pairing, and heartbeat monitoring — it never touches the actual workload or data - Authors keep 70% of the lease fee, paid monthly, gated on real heartbeat activity so an idle listing doesn't quietly earn money
Why this split matters to me: I didn't want to build a marketplace that requires trusting us to execute someone else's code on shared infrastructure. Keeping execution on the runner's own side removes an entire category of security/compliance conversation.
Right now it's early — prototype stage, a working control-plane dashboard, and 5 seed agents I built myself (PrestoDB EKS autoscaling, OpenStack Tempest testing, NFS Connectathon testing, a storage benchmarking agent, a REST API testing agent using LangGraph + CrewAI). Site's honest about being early — execution isn't live yet, it's a waitlist while I build out the payout/execution pipeline.
Would genuinely love feedback, especially from anyone who's tried to license or monetize internal tooling before, or who can poke holes in the revenue-share/heartbeat-gating model.
agent4lease.com