Skip to main content
As an orchestrator, you run a node that provides GPU compute to the Livepeer network. Orchestrators process workloads submitted by gateways - transcoding video, running AI inference, or executing BYOC containers - and earn LPT rewards and ETH fees in return. There are two approaches depending on whether you want full control of a single setup or prefer to separate rewards management from workload processing.

​
Pick Your Setup

Full go-livepeer Setup

Run the full go-livepeer binary on one machine - register on-chain, stake LPT, and process workloads directly.

Split Setup (Siphon + go-livepeer)

Separate rewards and keystore management from workload processing across two machines. Avoid missing rewards.

​
Option A: Full go-livepeer Setup

The standard path runs the full go-livepeer binary, registers on-chain, and actively processes workloads. Everything runs on one machine. This is the path for operators who want a straightforward, all-in-one orchestrator.

Install go-livepeer

Build from source or download a release binary. The go-livepeer CLI includes everything needed to run as an orchestrator - no additional tooling required.

go-livepeer Repository

Source code, releases, and build instructions.

Register and activate on-chain

Use the go-livepeer CLI to register as an orchestrator on the Livepeer protocol. You’ll need to stake LPT and activate your node on the Arbitrum L2.

go-livepeer Technical Docs

Docs for networking, GPU setup, payments, and Ethereum/Arbitrum configuration.

Configure your GPU and start processing

Set up your GPU configuration and start your orchestrator. Once active, your node receives workloads from gateways and processes them.

​
Option B: Split Setup (OrchestratorSiphon + go-livepeer)

The split setup separates two concerns across different machines:
  • Secure machine - runs OrchestratorSiphon, a lightweight Python toolkit that manages your orchestrator keystore. Handles on-chain actions like claiming rewards, voting on proposals, and updating your service URI. Your keystore stays on one secure, isolated machine.
  • GPU machine - runs go-livepeer to actively process workloads. No keystore needed on this box.
This avoids a common problem: missing rewards because your orchestrator node was busy processing workloads or went down temporarily. With the split setup, rewards claiming runs independently on its own machine.

OrchestratorSiphon

Lightweight keystore management - claim rewards, vote on proposals, update service URI. Runs on your secure machine.

go-livepeer

Deploy on your GPU machine for active workload processing. Point your service URI to this box.
You can start with OrchestratorSiphon alone as a passive orchestrator to earn rewards while you set up your GPU infrastructure. When you’re ready to process workloads, deploy go-livepeer on a separate machine and update your service URI to point to it.

​
Not sure which setup?

If you’re new to Livepeer and just want to contribute a GPU without running a full orchestrator, consider joining an existing pool instead. The Orchestrator Quickstart has a decision tree to help you choose.
Last modified on March 3, 2026