Full stack Bitcoin software

Client Compatibility

Reference documentation for each interface served by Bitcoin Server (bs): calling patterns, examples and responses.

The interfaces of bs are drop-ins for the interfaces of other servers, so that existing clients can use bs unchanged. These matrices show, for each client, the interfaces it uses and how it fares on each against bs: first the clients that have been run against bs, then those that use the same interfaces but have not yet been tested.

Works: the client did its job end to end. Partial: some of its features fail. Blocked: it fails. Retest: the failure is fixed in bs, and the client has not been run again. Expected: untested, but every method the client uses is served. No: the client requires methods on that interface that bs does not serve, such as bitcoind's wallet methods (see Node wallet clients). A blank cell is an interface the client does not use.

1. Tested#

Results are for the bs master branch on the date given, on mainnet. Each client links to its notes (§3).

Client Kind Version Date Native Electrum Sparrow JSON-RPC REST ZMQ BTCD Esplora P2P
lnd, bitcoind backend Lightning node 0.21.6 2026-11-07 Works Works
lnd, neutrino backend Lightning node 0.21.6 2026-11-07 Works
lnd, btcd backend Lightning node 0.21.6 2026-11-07 Partial
Sparrow Wallet Wallet 2026-10 Works Works
Specter Desktop Wallet 2.3.0 2026-11-16 Works No
SatSage Wallet analysis 1.0.1 2026-11-16 Blocked Works Expected Expected
kassiber Accounting 0.22.78 2026-11-06 Works No Blocked
rbitcoin-space Explorer 0.1.0 2026-11-16 Works Partial Partial
Fulcrum Electrum server 2.1.3 2026-10-16 Blocked Works
mempool Explorer 3.3.2 2026-11-17 Works Works
Bitcoin Core Peer 31.0 2026-10-17 Works

2. Untested#

Expected compatibility, from each client's source or documentation. It is not a test.

Client Kind Native Electrum Sparrow JSON-RPC REST ZMQ BTCD Esplora P2P
Electrum Wallet Expected
BlueWallet Wallet Expected
Nunchuk Wallet Expected
Stack Wallet Wallet Expected
Blockstream app Wallet Expected
BitBoxApp Hardware wallet app Expected
Trezor Suite Hardware wallet app Expected
Liana Wallet Expected No
Wasabi Wallet Wallet Expected Expected
Aqua Wallet Expected
Phoenix Lightning wallet Expected
Bitkit Lightning wallet Expected
Zeus Lightning wallet Expected
Blixt Lightning wallet Expected
Alby Hub Lightning node Expected
Core Lightning Lightning node Expected
btcwallet Wallet daemon Partial
LDK Node Lightning library Expected Expected Expected Expected
BDK Wallet library Expected Expected Expected Expected
Fedimint Federation Expected Expected Expected
BTCPay Server Payment processor Expected Expected
Ashigaru Dojo Wallet backend Expected Expected Expected
electrs Electrum server Expected Expected
BTC RPC Explorer Explorer Expected Expected
ord Index Expected
Floresta Utreexo node No

Custodial wallets, and wallets bound to their provider's backend, have no server setting and are not listed. Node platforms such as Umbrel, Start9 and RaspiBlitz host a node and its clients rather than connect to one.

Stratum v1 and v2 are planned for version 5. Their clients are mining devices and proxies, such as Bitaxe (AxeOS), cgminer, Braiins OS and the Stratum V2 reference translator.

3. Notes#

lnd#

bitcoind backend: wallet creation, newaddress (P2WKH, NP2WKH, P2TR), restart with addresses, graph sync and channel validation, and wallet recovery work. ZMQ blocks arrive within a second of validation.

neutrino backend: recovery of a wallet from 2019 (366,000 blocks) took about 15 minutes and found every reference transaction. Requires a store created with node.provide_filters.

btcd backend: the chain syncs, but once the wallet has an address, lnd calls the deprecated rescan with addresses, which bs does not serve, and the wallet never finishes syncing. Two client changes replace rescan with loadtxfilter and rescanblocks, as btcd recommends: btcsuite/btcwallet#1515 and lightningnetwork/lnd#11583, both open. With them, a 368,000-block wallet rescan found the same 397 transactions as the neutrino backend, and the lnd integration suite gives identical results against btcd with and without them.

Sparrow Wallet#

Wallets load, and silent payment wallets are scanned and notified over blockchain.silentpayments.subscribe. Requires the address index, and for silent payments node.silent_start_height.

Specter Desktop#

In Spectrum mode (Spectrum 0.8.0), Specter negotiates protocol 1.4. A watch-only BIP84 wallet imported by descriptor, with 600 receive and change addresses, synced 285 confirmed transactions, every one at the same height as a public Electrum server reports for the same addresses, and the same balance. Spectrum makes its requests one at a time, so its sync time grows with the round trip time to the server. In Core mode Specter keeps its wallets in the node, which requires bitcoind's wallet methods.

SatSage#

A tax and holding period analyzer for xpub wallets. Electrum: the analysis of an address with 126 unspent outputs gave the same outputs, amounts and origin traces as a public Electrum server, and dated when each output arrived, which that server left unknown. The history scan of a BIP84 wallet (198 addresses) gave 146 of 146 received outputs with the same values, heights and spending transactions as the public server.

Native REST: SatSage recognizes bs by libbitcoin in the server.version name, and would then read addresses, outputs, transactions and headers from the Native REST interface. It looks for address at the top level of configuration, which bs now reports under store, so it finds no address index and stays on Electrum, with the same results. With that check adjusted, its address, output and transaction queries all gave the same results as Electrum.

JSON-RPC (scantxoutset, getrawtransaction) and client filters over P2P were not tested. Its import of xpubs from a node wallet (listdescriptors) requires a node wallet.

kassiber#

Electrum: 436 of 436 transactions identical to a public Fulcrum server, with amounts, fees and timestamps. Esplora: a client fault. The client computes script hashes in Electrum byte order, reversed, for Esplora's /scripthash/ paths, which take them unreversed, so it finds no transactions on any Esplora server. JSON-RPC: requires bitcoind wallet methods (createwallet, importdescriptors, listwallets).

rbitcoin-space#

Blocks, transactions, outspends, fee estimates, mining pool and hash rate views, and address pages work. The fee recommendation and memory pool views fail: getmempoolentry lacks chunkweight (bitcoind 31), getmempoolcluster awaits the version 5 pool (-33), and getmempoolfeeratediagram is not served (-32601). vsize_bip141 and getblocktemplate are now served, and have not been run again. An Esplora address with a very long history is served in full, which can take minutes, and the explorer's other address requests wait meanwhile.

Fulcrum#

Fulcrum requires a Content-Length header on each JSON-RPC response, and bs sends single responses with chunked transfer encoding, which HTTP/1.1 allows. With a test build that sized single responses, Fulcrum indexed half of mainnet (to height 483,000) at about 9 blocks a second, with ZMQ block notifications. bs serves the Electrum interface itself, so Fulcrum is not needed in front of it.

mempool#

The mempool backend in its electrum mode, which reads blocks, transactions and the pool over JSON-RPC and addresses over Electrum. It indexed the latest blocks, followed new ones, loaded the pool and served fee recommendations. Against mempool.space: block headers, transaction lists, transactions, their hex and merkle proofs, the confirmation status of transactions, fee recommendations and the 126 unspent outputs of an address all identical.

The differences from mempool.space are those of this backend mode, and the same with Bitcoin Core: spent outputs without the spending transaction, address totals without output counts, address histories in pages of 10, and no block status. The pool held fewer transactions than mempool.space, which reflects the pool of the node, not the interface.

The backend makes one request per transaction, and per input for its previous output, one after another, so it should run close to the node: at a round trip of 186 ms one block took several minutes, on the same host about a second.

Bitcoin Core#

A differential test fed the same blocks, valid and invalid, to bs and to Bitcoin Core over P2P: 15 of 15 decisions identical. Reorganization tests against Bitcoin Core on regtest (depths 1 to 1000, ties, double spends, restarts mid-reorganization): 35 of 35 passed.

Untested clients#

Electrum wallets negotiate protocol 1.4 to 1.6, and bs serves 1.0 through 1.7. Core Lightning calls bitcoin-cli for getnetworkinfo (version 23.0 or later, localrelay), getblockchaininfo, getblockhash, getblock (0), estimatesmartfee, getmempoolinfo, gettxout, sendrawtransaction and getpeerinfo, and getblockfrompeer only for pruned nodes. electrs refuses a node that reports pruned. Wasabi, Zeus, Blixt and BDK (Kyoto) require client filters. LDK Node reads blocks over REST with JSON-RPC for the rest. Bitkit takes an Electrum server over TCP or TLS. Aqua takes one for bitcoin only over TLS, with a certificate valid for the host name. Alby Hub's built-in node (LDK) takes an Esplora server (LDK_ESPLORA_SERVER), and Alby Hub can instead use an lnd or Core Lightning node. ord's index and server are expected to work, and ord wallet requires a node wallet. Floresta requires Utreexo proofs from its peers.

Node wallet clients#

Some clients keep their wallet in the node, and call bitcoind's wallet methods to create it, import descriptors or addresses, list and lock unspent outputs, fund, sign and send. bs does not implement the wallet methods (see JSON-RPC), and these clients have no other backend, so they cannot be run on bs:

Client Kind Wallet methods required
Eclair Lightning node fundrawtransaction, walletprocesspsbt, listunspent, lockunspent, getnewaddress, sendtoaddress, …
JoinMarket Coinjoin importmulti (legacy wallet), or importdescriptors in JoinMarket-NG, and listunspent

Liana, Specter Desktop and kassiber also need a node wallet on JSON-RPC, and work over Electrum instead, so they stay in the matrices with No on JSON-RPC.

Floresta's No is unrelated: it requires Utreexo proofs from its P2P peers, which bs does not provide.

Copyright © 2011-2026 libbitcoin developers