Bittensor subnet 105
Bandwidth you can measure
Beam is an open coordination layer for data transfer. Describe where bytes need to go and a competing network of operators moves them — measured, verified, and paid on delivered performance.
Pick your interface
Start from the map, build visually in Studio, or manage your account from a terminal. Every transfer your organization creates spends the same credits and shows up together in the Console.
Build with BeamStart hereWhat a client needs end to end: credits, an API key, the interfaces to choose between, and the storage to connect.Read the guideBeam StudioBuild it visuallySelf-host the app that turns transfers, transforms, checks and schedules into a graph you can watch run.Read the guideBeam CLIManage it from a terminalInstall the CLI, sign in, and manage your account, the machines connected to your organization, and Rooms.Read the guide
Start in five minutes
Create an API key, install what you will drive Beam with, then send bytes. The SDK handles chunking, assignment, and recovery.
python -m pip install beam-network-sdk
async with BeamSDK(api_key=..., environment="prod") as beam:
transfer = await beam.transfers.create(
sources=[SourceConfig(type="http", url=source)],
destinations=[DestConfig(type="http", url=destination)],
total_size=size,
)
await beam.transfers.distribute(transfer.transfer_id)
await beam.transfers.wait_complete(transfer.transfer_id, timeout=3600)- Add creditsFund your organization in the Console under Settings → Billing.
- Create an API keyScope it to a project, and keep it out of source control.
- Install the SDK
pip install "beam-network-sdk[s3]", or the TypeScript, Go and Rust SDKs. - Install Beam Studio
curl -fsSL https://cdn.b1m.ai/studio/install.sh | sudo sh— the self-hosted app for building data movement visually. Needs a Linux host with systemd, and installs Docker itself; see the requirements. - Install the Beam CLI
curl -fsSL https://cdn.b1m.ai/cli/install.sh | sh— manage your account, machines, and Rooms from a terminal. - Send a transferPoint the SDK at a source and a destination, then wait for completion — or run it from Studio, without writing code.