Put your trusted computers to work together in one private loop over your own network — compatible compute, direct file sharing and automatic device discovery. See what each machine can contribute. Keep control of your loop.
One installer on every machine. Devices find each other, pair once, then work as a team.
Compatible workloads can be split across participating machines and their results verified. Run a benchmark to measure what your own loop contributes.
See available memory on each device. Ramster can use that information to place compatible work. Memory remains physically on each computer.
The engine checks available resources to help place compatible tasks. The live loop view shows reported devices, queued work and active jobs as they change.
Browse shared files on connected devices, upload and download, or send a file directly to a peer over your local network.
Devices pair explicitly and every peer call authenticates with per-pair tokens. Your loop stays yours — revocable at any time.
See task activity and benchmark results. Energy, carbon and cloud-cost equivalents are estimates, separate from measured work.
Ramster is an engine for moving compatible work and files between the machines you already own.
Open Ramster on compatible computers on the same network. Local discovery looks for other running Ramster devices.
Pair trusted devices explicitly. Review connected devices and revoke access when needed.
Browse another connected machine's shared files, send a file, or download what you need directly over the local network.
Request a Ramster source ZIP from a running peer. After it accepts, download the package through Files. Get the Windows installer from the download page.
Your devices, your network, your choice of peers. Local file sharing does not require a cloud-storage account.
Use your local network to move files and compatible work between connected computers. You choose which devices participate.
Discovery, pairing, compute and file transfer all happen inside your own network. Unplug the internet and Ramster keeps working.
Pairing uses per-device tokens; scheduling learns from aggregate outcomes like speed and reliability — never task contents.
Remove any device and it drops out of the loop instantly — its access tokens stop working on the spot.
For the engineers: this is exactly what happens when you press go.
UDP broadcast DISCOVER / PEER_INFO sweeps find every node on the subnet. Responses are deduplicated per sweep, remembered across sweeps, and sweep timing adapts to how responsive the network actually is.
Explicit pairing issues per-device credentials; every peer call carries bearer + device-identity headers. Strict-trust mode rejects unpaired callers outright.
A job is split into shares proportional to cpu_cores × log2(RAM), blended with measured hash-rate benchmarks, idle-state weighting and self-learned reliability weights — so the fastest free machine always gets the biggest slice.
All peer shares fire simultaneously while local work runs in parallel — wall-clock approaches the slowest node, not the sum. Adaptive per-peer timeouts, transient retry, and automatic local re-execution help recover from failed work.
Every node publishes a file manifest; the loop merges them so you can browse and pull files from any device as if they were local.
In our scaling benchmark, 100 simulated nodes connected and shared file manifests in a single loop with ~1.1 s full fan-out and flat scheduler memory (~57 MB). On home WiFi the practical ceiling is usually the router's subnet size (~250 addresses) — not Ramster.
Start with what your loop can do today.
Send project files between connected computers without first uploading them to cloud storage. Browse and download files shared by your peers.
Run test compute and benchmarks to see how participating machines handle supported workloads. Ordinary applications do not automatically become distributed.
Inspect real device resources, task activity and available benchmark results. Measure your own loop before deciding where work belongs.
Built toward a future where the hardware you already own can work as one intelligent local fabric. That is the direction; the capabilities above are today's starting point.
Every task follows the same disciplined path across the loop.
Ramster is in closed early access. The download is sealed — enter your access code to break the seal.