2Node
Sync the network yourself.
fiddy node up installs Docker and the official Snapchain bootstrap on your server, then starts syncing. You now run a read node: it follows the chain, stores the full state, and serves HTTP on :3381 and gRPC on :3383. Anyone can run one. No permission, no API keys.
Want an agent to drive instead? Use the agent hand-off at the top. Its linked guide is written for agents, with the steps, prices, and mistakes to avoid.
On Linux the CLI installs the Docker engine itself through get.docker.com, adds you to the docker group, and asks you to log back in once. There is no Docker Desktop on a headless box. On macOS it offers Docker Desktop instead. The CLI is bin/fiddy: run export PATH="$PWD/bin:$PATH" after cloning, or every command says fiddy: command not found.
fiddy node up # install, start, wait for the API to answer
fiddy node status # sync progress
fiddy node logs # tail the node
fiddy node dashboard # Grafana at localhost:3000 through an SSH tunnel
fiddy node autoupgrade on # weekly upgrades, Tuesday 03:00
Autoupgrade is off by default. The upstream job ends with a host-wide Docker prune. That is fine on a dedicated box, never on a shared one.
Read nodes vs validators, and where your casts go
Validators propose and vote on blocks. Only keys in the validator set can. Read nodes do everything else, including accepting writes.
When you submitMessage to your read node, it validates the message and gossips it to validators. A read node has no mempool of its own. A success response means “valid, accepted for relay,” not “in a block.” Confirm inclusion by watching for a MERGE_MESSAGE event or querying the message back.
Under heavy network load a gossiped message can drop and need resubmitting. The client can fall back to submitting at a validator endpoint, and this tutorial shows the trade-off when it matters.
Downloading the whole history: 5,690 pieces, about 468 GB total. At the measured 10-20 MB/s, this takes many hours. Plan for overnight on a slow disk.
0%
Your computer is downloading the whole history of what everyone has ever posted. No exact finish time yet. At 10-20 MB/s this takes many hours.
You can close this and come back. It keeps going on its own.
34%
About a third is down. It is pulling 10-20 megabytes every second. Roughly 300 GB is left. The speed depends on your connection and disk, so plan for overnight.
You can close this and come back. It keeps going on its own.
99%
The download finished. Now it unpacks everything into working space. That step needs the full 2 TB free at once, then it shrinks to about 597 GB. A few more minutes.
You can close this and come back. You're nearly there.
Your copy will not answer any requests until the whole history is downloaded and unpacked. That is normal, not broken.
If your copy is not answering yet
A brand-new copy refuses every request until the whole history finishes downloading and unpacking, which takes hours. Run fiddy node logs to watch progress, then keep waiting; fiddy node status works after the load completes.
If your disk filled up
The first boot briefly needs about 2 TB, with a measured peak of about 1.66 TB. A smaller disk fills and the whole download starts over from chunk 1. Free up space or move to a computer with 2 TB free, then run it again. Do not retry on the same full disk.
If your laptop went to sleep
Your computer paused, so a local download paused with it. Keep it plugged in and awake, or move to a computer that stays on (about $120-285/month to rent one).
