Cardano Client Lib · native · offline
One battle-tested SDK.
Four languages.
No JVM.
Cardano Client Lib compiled to a native shared library with GraalVM. Offline key derivation, transaction building, signing, Plutus, and governance — from Python, Go, Rust, and JavaScript.
Same semantics in every language
Same API groups, same error codes, same TxPlan YAML — kept at parity and enforced by CI.
import { CclBridge, TESTNET } from '@bloxbean/cardano-client-lib';
const bridge = new CclBridge();
const account = bridge.account.create(TESTNET);
const result = bridge.quicktx.build(yaml, utxos, params);
const signed = bridge.account.signTx(mnemonic, TESTNET, 0, 0, result.tx_cbor);
// fully offline — submit signed cbor with any HTTP clientbridge, _ := ccl.New()
defer bridge.Close()
account, _ := bridge.Account.Create(ccl.Testnet)
result, _ := bridge.QuickTx.Build(yaml, utxos, params)
signed, _ := bridge.Account.SignTx(mnemonic, ccl.Testnet, 0, 0, result.TxCbor)
// cross-compiles: pure Go, no C toolchainlet bridge = ccl::Bridge::new()?;
let account = bridge.account().create(ccl::network::TESTNET)?;
let result = bridge.quicktx().build(&yaml, &utxos, ¶ms)?;
let signed = bridge.account()
.sign_tx(&mnemonic, ccl::network::TESTNET, 0, 0, &result.tx_cbor)?;
// native lib fetched once at first build, then cachedfrom ccl import CclLib, Network
with CclLib() as lib:
account = lib.account.create(Network.TESTNET)
result = lib.quicktx.build(yaml, utxos, params)
signed = lib.account.sign_tx(mnemonic, result["tx_cbor"], Network.TESTNET)
# wheels bundle the native library — pip install and goWhy these bindings
No JVM at runtime
Cardano Client Lib compiled to a native shared library with GraalVM native-image. Native startup, small footprint, a plain C ABI.
Fully offline by design
Key derivation, transaction building, signing, Plutus data — no network calls, no submission, no secrets in the library. You bring chain data; optional providers fetch it for you.
One YAML transaction format
Describe payments, staking, governance, minting, and Plutus operations as TxPlan YAML — identical across all four languages, with a verified intent catalog.
Offline Plutus costing
The embedded Scalus UPLC evaluator computes execution units in-process. Override with explicit units or a remote evaluator when you want node-backed costing.
Four first-class wrappers
Python, Go, Rust, and JavaScript (Bun) are kept at strict parity — same API groups, same error codes, enforced by CI. No second-class language.
AI-friendly from day one
llms.txt, a full-docs dump, and an AI Starter Pack distilled for agents: point Claude Code or Cursor at one URL and generate correct code immediately.
Where this fits
Cardano already has excellent native libraries — pycardano, MeshJS and Lucid Evolution, pallas, gOuroboros, Apollo, and more. If one of them serves you well, use it. These bindings shine when it doesn't quite: when a native library is missing a capability you need, or its API doesn't fit how you work, you get Cardano Client Lib's exact, well-tested semantics — identically in four languages — without leaving your stack. Being offline and stateless, they slot in alongside an existing library for just the missing pieces.
Choose these bindings when…
- your native library is missing functionality you need — e.g. Conway governance operations, offline Plutus costing, HD-wallet or DRep key derivation — use these bindings for just those pieces, alongside it
- you're not happy with an API and prefer CCL's model — declarative TxPlan YAML instead of imperative builders
- you want identical transaction-building behavior across several languages or services, or already use CCL on the JVM
- you want offline building backed by CCL's test suite and devnet-verified intent shapes, and insurance against a library going unmaintained
Choose a native library when…
- you prefer a pure-language dependency — these bindings ship a ~50 MB native binary
- you need node protocols / chain sync — deliberately out of scope here
- your platform isn't covered (e.g. macOS Intel)
- an idiomatic library already fits your stack — no need to switch
Bring your AI agent
The docsite ships llms.txt, a full-docs dump, and a distilled
Starter Pack. One curl and Claude Code, Cursor, or any agent
writes correct code against the bindings.
$ curl -o CLAUDE.md https://pages.bloxbean.com/cardano-client-bindings/ai/starter-pack.md