Cross-chain
Every chain keeps its own terms, but they all land in one database with the same shared fields: id, time, sender, receiver, amount, fee, status, chain and network. That is what lets one command look at all of them at once.
Metrics side by side
--chain all reads every chain and network that has stored transactions:
multichain-indexer --chain all metrics --days 365
Chain Transactions Per day Active Fees Fees (USD) Failed
algorand testnet 108 0.3 6 0.108 ALGO n/a 0.0%
bitcoin testnet4 4 0.01 4 0.00001472 BTC n/a 0.0%
ethereum testnet 19 0.05 1 0.026567220971150023 ETH n/a 0.0%
near testnet 20 0.05 2 0.0221832130514401 NEAR n/a 25.0%
solana devnet 6 0.02 2 0.00003 SOL n/a 0.0%
Last 365 days. US dollars only for mainnet, at the price in config.toml.
This came from a database that version 0.6.0 wrote while syncing one small account on each
test network. --json prints the same rows as JSON, with each chain's transactions per day.
A chain that cannot be read, for example because the plugin that added it is no longer installed, shows its error on its own row while the others are still reported.
Fees in US dollars
Fees are always shown in each chain's own unit. A US dollar figure appears only for mainnet
records, and only when you give a price, with its date, in config.toml:
[prices.bitcoin]
usd = 98000.5
date = 2026-11-15
Nothing is fetched from a price service, so the same database and the same file always give
the same report. Test network tokens have no market price, so test networks show n/a.
The report
multichain-indexer --chain all report --days 365 --output report.html
writes one HTML file with a summary table and charts of transactions, active accounts, fees
in US dollars when priced, and transactions per day for each chain. It has no scripts and
loads nothing, so it opens offline and can be attached to an email. --format csv and
--format json write the same rows as data.
An example report was made from the database above.
One export for every chain
multichain-indexer --chain all export --format parquet --output all.parquet
writes every chain's transactions into one table: the shared fields as columns, amounts and
fees as text so values past 64 bits stay exact, and each chain's native block as a JSON
column. It loads into one pandas DataFrame:
import pandas as pd
frame = pd.read_parquet("all.parquet")
frame.groupby(["chain", "network"]).size()
An export never leaves out a chain silently: if one cannot be read, the export stops and names it.
Syncing many accounts
List the accounts in config.toml, by chain and network:
[accounts.bitcoin]
testnet4 = ["mgdKMwnT4S5gbx55Uw3tRJchUSrHwAFjzu"]
[accounts.algorand]
testnet = ["JCN7TCSQEG6YHAQ7F6BS73J2UDXI2L7S7WJU35LYMIK22IL264ZRJO55QM"]
multichain-indexer --chain all sync
Chains are synced in parallel and the accounts of one chain one after another. Mainnet
accounts are synced only with --network mainnet. One failing account never stops the others;
each is reported, and the command exits with 1 if any failed.
Free API plans limit a key, not a chain: Blockscout serves Ethereum, Arbitrum and OP Mainnet on one key. So every chain that uses one key, for Blockscout, Etherscan, NodeReal or Alchemy, shares one rate limiter, and every keyless endpoint is limited per host. With one limiter per chain, three chains on one Blockscout key at 4 requests a second each drew HTTP 429 on 26 to 47 of 90 requests; with one shared limiter, 0 or 1.
Following an account
multichain-indexer --chain bitcoin sync mgdKMwnT4S5gbx55Uw3tRJchUSrHwAFjzu --follow
runs the ordinary incremental sync again and again, storing new transactions as they arrive,
until Ctrl-C, which prints a summary. A round that fails on the network is tried again on the
next round. The pause between rounds suits each chain's block time and the cost of asking a
public endpoint for its head; --interval changes it.
| Chain | Default pause | Requests to read the head |
|---|---|---|
| Algorand | 10 s | 1 |
| EVM chains | 15 s | 2 |
| Solana | 30 s | 1, at 1 request every 2 seconds |
| NEAR | 30 s | 2 |
| Bitcoin | 60 s | 1 |
| Plugin chains | 30 s |
config.toml
config.toml holds endpoints, default networks, accounts and prices. It is read from the
current folder, or from --config. It sits below .env and the environment, which win.
[settings]
requests_per_second = 2
[chains.bitcoin]
network = "signet"
[chains.bitcoin.signet]
esplora_server = "https://my-esplora.example/signet/api"
[chains.<chain>.<network>] entries become the same settings .env takes, here
BITCOIN_SIGNET_ESPLORA_SERVER. A chain's network becomes its default, but mainnet is
always chosen with --network mainnet on the command line: network = "mainnet" in
config.toml is refused with an error that says to pass --network mainnet instead. The file is meant to be shared, so it never holds
secrets: a key ending in _API_KEY or _TOKEN is refused with an error that names the key
and never shows its value. Keys, and a database URL with a password, belong in .env.
From Python
from pathlib import Path
from multichain_indexer import CrossChain
with CrossChain() as everything:
for outcome in everything.sync():
print(outcome.chain, outcome.account, outcome.error or outcome.result.records_fetched)
report = everything.metrics(days=30)
for row in report.rows:
print(row.chain, row.network, row.transactions, row.total_fees_display)
everything.export("parquet", Path("all.parquet"))
AsyncCrossChain has the same methods as coroutines. Indexer.follow and
AsyncIndexer.follow follow one account.