Matrix
Most chains offer Full only. Bitcoin and BSC also offer Pruned, which trims disk by dropping old block history at the cost of not serving the full history; Tron offers Lite, which starts from a recent block instead of carrying deep history; Stable offers Pruned and Archive (all historical state) rather than Full.
A multi-role chain runs more than one process per node — Ethereum runs an execution client (
el) plus a consensus client (cl); Igra runs a kaspad node plus an igra-reth execution layer; Ink runs an op-reth execution layer plus an op-node rollup client; Monad runs three. Each role is a separate process (Agent) or container (Operator) with its own resources, and you pick its client and version independently in the wizard.
Igra and Ink are container-only — their clients ship as images, not binaries, so they run as pods on an Operator executor and as Docker containers on a bare-metal Agent, where the wizard’s Runtime field locks to Docker. Most other chains run as systemd services on an Agent, though Ethereum and Tempo can also run as Docker containers there (pick the runtime in the wizard). HyperLiquid, Tempo, and Stable are Agent-only — not offered on Kubernetes (HyperLiquid additionally runs one per Agent).
Resource defaults
The deploy wizard pre-fills the resource form with per-network, per-client defaults sized for a fresh sync. You can change them; if your selection exceeds the executor host’s available capacity, the wizard shows a non-blocking warning. The numbers below are for Full nodes. For single-process chains the defaults are fixed per network:
Disk above is for Full nodes; Bitcoin Pruned, BSC Pruned, and Tron Lite trim the disk request, and Stable is listed per node type (Pruned / Archive) since it has no Full — same figures on either network (see their per-chain pages).
For multi-role chains — Ethereum, Igra, and Monad — each role is sized separately, and the total is the sum of its roles’ requests. The wizard shows the live numbers for your exact selection as you choose clients, so deploy there to see the current defaults.
What every chain shares
Whatever client you choose, these conventions hold:- Outbound only. No inbound ports need exposing for the platform to manage your node. You can still choose to expose RPC for your own use.
- One data directory per node. Config and data live under a single mount (Operator: one volume; Agent: a subdirectory under
data_dir). Monad is the partial exception — its triedb lives on its own volume. - Setup is handled for you. JWT secret generation (Ethereum, Igra), snapshot bootstraps where a chain supports them, and checkpoint sync URLs (Ethereum consensus client) all run as part of node startup — they aren’t config you maintain.
- Guarded overrides. Each chain marks some settings as managed (rejected if you try to change them, because the platform owns them) and exposes the rest as typed, described keys with hints in the deploy wizard. See Edit a running node.