One workflow, two trust models
Compute moves through the network in four clear steps.
PlanetNodes sends each supported job to a suitable private-cluster machine or marketplace provider. It does not add remote RAM or GPU power directly to your laptop.
Machines you trust.
Invite computers owned by you, your school, lab, or company into a private pool. The scheduler distributes independent tasks among them.
Explore Cluster →Capacity from outside providers.
Choose an independent provider by capability, reputation, region, and price. Use it only for portable workloads appropriate for that trust boundary.
Explore Marketplace →Your 16 GB laptop cannot run a 32 GB job.
Rather than replacing the laptop, choose a provider computer with at least 32 GB of memory and submit the supported workload to that machine.
The provider runs the job remotely. PlanetNodes returns the output, logs, and artifacts. Your laptop still has 16 GB of RAM throughout the process.
PlanetNodes is best for portable batch work, AI inference, builds, tests, and other jobs that can run away from your local desktop.
- 01
Submit a job
A customer defines the workload, region, budget, and whether the result needs independent verification.
- 02
Match capacity
The scheduler selects an eligible online node using capability, availability, price, reliability, and policy constraints.
- 03
Execute securely
The provider agent accepts the signed assignment, runs it within configured limits, and reports status and artifacts.
- 04
Verify and settle
PlanetNodes checks the execution receipt, compares redundant results when requested, and settles completed work.
Two sides of one job
What can the provider see?
The provider does not receive control of your laptop or watch your screen. But they own the computer running the job, so they may be able to inspect the submitted input, source files, model data, process memory, logs, and output.
What is removed afterward?
The supported agent deletes the temporary Docker container, cloned workspace, and output archive after the result is uploaded. This is best-effort cleanup, not secure erasure. Cached model downloads may remain for later jobs, and a host owner could retain copies.
Choose your side of the network.
Submit compute work or contribute hardware through the provider flow.