Find which derivation path a wallet used

Everything below runs inside the sealed container through sealed-run (the agent sees addresses and paths only) or, for a human at the terminal, via ./blocktools reveal. Flags carry public parameters only; the seed phrase arrives on stdin.

1. See the plan first (no keys involved)

derive --explain --preset recovery --chain all
derive --list-presets

2. "Which path produces this address?"

Put the public addresses you know the wallet had in targets.txt, one per line:

seed-repair < /in/seed.txt | seed-to-master | derive --preset exhaustive --chain all --match /in/targets.txt

Only hits are printed, each with chain, kind and path. If nothing matches, {"status":"no_candidates","count":N} says how many derivations were tried.

3. Common one-liners

Want	Flags
First 200 native-segwit receive + change	--chain btc --address-type p2wpkh --change 0,1 --index 0..199
A custom path	--chain btc --path "m/0'/0'/{i}'" --both
BIP44 paths encoded as bech32	--chain btc --purpose 44 --address-type p2wpkh
Litecoin with a non-standard coin type	--chain ltc --coin-type 0
Ledger Live ETH accounts 0..9	--chain eth --preset ledger-live --index 0..9
TronLink	--chain tron --preset tronlink
Public addresses only (no key field in the JSON)	--public-only
Solana only	--curve ed25519 --chain sol
Keep scanning while hits continue	--match targets.txt --gap-limit 20
Stop at first known address	--match targets.txt --stop-on-match
Uncompressed (2011-era) keys	--chain btc --address-type p2pkh --uncompressed

Ranges are inclusive: 0..9, 2-4, 0,1,5, 0..2,7.

4. A plain address list (SwapChan SCAN_JS layout)

--preset scan-js is the same wallet path table as SwapChan's SCAN_JS script: the same wallet names, paths and order, 20 addresses each, over eth, tron, sol, btc, ltc, doge, bch and btg. --address-list prints one label<TAB>path<TAB>address line per address and no key material (status lines go to stderr). To also try passphrases, put one per line in a file:

./blocktools run 'seed-repair < /in/seed.txt | seed-to-master --passphrase-file /in/passphrases.txt | derive --preset scan-js --chain btc --address-list'

The empty passphrase is tried first, then each non-empty line; rows from line N get [passphrase #N] after the path, as SCAN_JS prints them. Without --chain the preset uses all eight chains above; with --chain it keeps only the rows for those chains.

Through the sealed boundary this works as is, so addresses.txt can be written on the host:

./blocktools run 'seed-repair < /in/seed.txt | seed-to-master | derive --preset scan-js --address-list'

(add > addresses.txt on the host to keep the list).

The boundary passes a non-JSON line only if it is exactly label<TAB>path<TAB>address with a short label, a BIP32 path (optionally [passphrase #N]) and an address, and it passes the same leak check as JSON lines; any other line is dropped. The last line is the boundary's own {"tool":"redact",...} count; strip it with grep -v '"tool":"redact"'.

5. A typed BIP39 passphrase

A BIP39 passphrase is typed with seed-to-master --passphrase-prompt (built: it reads /dev/tty, not echoed, empty = none; checked on a pty 2026-10-26). It cannot be combined with --passphrase-file or a stdin passphrase.

Electrum 1.x old seed

seed-to-master prints such a seed as "scheme":"electrum-old" (hex seed, no xprv); derive walks the non-BIP32 keys m/{change}/{i} (change 0 and 1), uncompressed p2pkh, btc only. --preset electrum-old names the same plan and refuses an xprv line (unsupported). See input-electrum.md.

Planned (not built yet)

• TON addresses from derive (derive --chain ton gives no_candidates).

Tested examples

--public-only keeps the JSON rows but drops the key field entirely (not even [REDACTED]):

./blocktools run 'seed-repair < /in/seed.txt | seed-to-master | derive --chain btc --count 1 --public-only'


./blocktools run 'seed-repair < /in/seed.txt | seed-to-master | derive --preset exhaustive --chain btc --match /in/targets.txt'


./blocktools run 'derive --list-presets'