xrpl-mithril-codec 0.3.3

A next-generation, pure Rust SDK for the XRP Ledger.

xrpl-mithril is a Rust-native SDK targeting the 2026 XRPL protocol surface -- rippled v3.1.0 and beyond. It covers 50+ transaction types including Multi-Purpose Tokens, Token Escrow (XLS-85), AMM, Credentials, DynamicNFT, and every other mainnet feature through February 2026. The entire codebase enforces #![forbid(unsafe_code)] and builds on any platform without a C toolchain. Swap in the native libsecp256k1 backend with a single feature flag when you need maximum throughput.

Quick Start

Send 10 XRP on testnet:

use xrpl_mithril::client::JsonRpcClient;
use xrpl_mithril::tx::builder::PaymentBuilder;
use xrpl_mithril::tx::autofill::autofill;
use xrpl_mithril::tx::{sign_transaction, submit_and_wait};
use xrpl_mithril::types::{Amount, XrpAmount};
use xrpl_mithril::wallet::{Algorithm, Wallet};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // 1. Generate wallets
    let sender = Wallet::generate(Algorithm::Ed25519)?;
    let receiver = Wallet::generate(Algorithm::Secp256k1)?;

    // 2. Build a Payment transaction
    let mut unsigned = PaymentBuilder::new()
        .account(*sender.account_id())
        .destination(*receiver.account_id())
        .amount(Amount::Xrp(XrpAmount::from_drops(10_000_000)?)) // 10 XRP
        .build()?;

    // 3. Connect and autofill (fee, sequence, last ledger sequence)
    let client = JsonRpcClient::new("https://s.altnet.rippletest.net:51234")?;
    autofill(&client, &mut unsigned).await?;

    // 4. Sign -- produces a SignedTransaction (can't submit unsigned!)
    let signed = sign_transaction(&unsigned, &sender)?;

    // 5. Submit and wait for validation
    let result = submit_and_wait(&client, &signed).await?;
    println!("Validated in ledger {}: {}", result.ledger_index, result.result_code);
    Ok(())
}

See 17 runnable examples covering payments, escrow, AMM, MPTs, NFTs, multi-sign, credentials, and more.

Installation

Default (pure Rust, no C toolchain required):

[dependencies]
xrpl-mithril = "0.5.3"

With native cryptography backend (for maximum secp256k1 performance):

[dependencies]
xrpl-mithril = { version = "0.5.3", features = ["native-crypto"] }

By default, xrpl-mithril uses pure Rust cryptography (k256 + ed25519-dalek). The native-crypto feature swaps in libsecp256k1 via the secp256k1 crate for ~2x faster ECDSA signing and verification. Both backends expose the identical API -- switching is a Cargo.toml change, not a code change.

Individual Crates

For a smaller dependency footprint, depend on the crates you need:

Why xrpl-mithril?

1. Type-Level Safety

UnsignedTransaction<T> and SignedTransaction<T> enforce the build-sign-submit workflow at compile time. Submitting an unsigned transaction is not a runtime error -- it is a compile error. The method does not exist on the type.

2. Full 2026 Protocol Coverage

Targets rippled v3.1.0+. Covers Multi-Purpose Tokens (MPTs), Token Escrow for all asset types (XLS-85), DynamicNFT, AMM (XLS-30), Credentials, Price Oracles, and 50+ transaction types. Not a port of an older SDK -- researched from the rippled source and XRPL specification directly.

3. #![forbid(unsafe_code)]

Zero unsafe blocks in the entire codebase. Enforced at the workspace level via [workspace.lints.rust] unsafe_code = "forbid". Dependencies may use unsafe internally (they are audited by their communities), but every line of xrpl-mithril code is provably safe.

4. Pure Rust by Default, Native by Choice

Builds on any platform with cargo build. No C compiler, no OpenSSL, no system dependencies. TLS via rustls, cryptography via RustCrypto. Opt into libsecp256k1 when you need the throughput.

5. Transport-Agnostic Client

The Client trait abstracts over JSON-RPC and WebSocket transports. Write your application logic once, swap transports without changing code. WebSocket subscriptions provide real-time ledger streams.

6. Rust 2024 Edition

Edition 2024, MSRV 1.85. Uses let_chains, async closures, and the latest language features. This is not legacy Rust code getting maintained -- it is written for the Rust of today.

Architecture

┌─────────────────────────────────────────────────────────┐
│                     xrpl-mithril                        │
│                (facade — re-exports all)                │
└──┬──────┬──────┬───────┬───────┬───────┬────────────────┘
   │      │      │       │       │       │
┌──▼──┐┌──▼──┐┌──▼───┐┌──▼───┐┌──▼────┐┌─▼──┐
│types││codec││models││wallet││client ││ tx │
│     ││     ││      ││      ││       ││    │
│     ││     ││      ││ k256 ││reqwest││    │
│     ││     ││      ││  or  ││  +    ││    │
│     ││     ││      ││secp- ││tungst-││    │
│     ││     ││      ││256k1 ││enite  ││    │
└──┬──┘└──┬──┘└──┬───┘└──┬───┘└──┬────┘└─┬──┘
   │      │      │       │       │       │
   │   ┌──┘      │       │       │       │
   │   │ depends │       │       │       │
   │   │ on types│       │       │       │
   └───┴─────────┘       │       │       │
                         │       │       │
              ┌──────────┴───────┴───────┘
              │ tx depends on
              │ client, wallet, models, codec
              └──────────────────────────────

~15,750 lines of library code across 7 crates, plus ~3,720 lines of examples across 17 runnable demos. Every crate enforces #![forbid(unsafe_code)] and clippy::unwrap_used = "deny".

Feature Comparison

xrpl-mithril is not a fork or port of any existing SDK. It was designed from scratch by studying the rippled source code and XRPL protocol specification to provide idiomatic Rust coverage of the 2026 protocol surface.

Examples

All examples connect to XRPL testnet and require network access. Fund test wallets via the testnet faucet.

WebSocket Subscriptions

use futures::StreamExt;
use xrpl_mithril::client::{Client, WebSocketClient};
use xrpl_mithril::models::requests::subscription::SubscribeRequest;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let client = WebSocketClient::connect("wss://s.altnet.rippletest.net:51233").await?;
    let mut stream = client.subscribe_stream()?;

    client.request(SubscribeRequest {
        streams: Some(vec!["ledger".to_string()]),
        accounts: None,
        accounts_proposed: None,
        books: None,
    }).await?;

    while let Some(msg) = stream.next().await {
        if msg["type"].as_str() == Some("ledgerClosed") {
            println!("Ledger #{}: {} txns",
                msg["ledger_index"], msg["txn_count"]);
        }
    }
    Ok(())
}

Design Philosophy

• Invalid states are unrepresentable. IssuedValue stores mantissa and exponent matching the XRPL wire format exactly. Construction validates invariants. Every instance in memory is guaranteed valid.

• Strong newtypes prevent type confusion. AccountId, Hash256, CurrencyCode, and Blob are all distinct types wrapping byte arrays. Passing a hash where an account ID is expected is a compile error, not a protocol-level bug discovered in production.

• #[non_exhaustive] enums for protocol evolution. The Transaction and TransactionType enums use #[non_exhaustive], so new XRPL amendments can add transaction types without breaking downstream code. Unknown types are preserved in a catch-all Unknown variant.

• Zero-cost cryptography abstraction. The signing layer is generic over the cryptography backend via the ecdsa and signature trait crates. Backend selection is resolved at compile time with no runtime dispatch.

• Architecture decisions are documented. See research/design-decisions.md for 6 ADRs covering decimal representation, transaction enum strategy, newtype strategy, signed vs. unsigned transaction types, field definitions, and cryptography backend selection.

Protocol Coverage

Minimum Supported Rust Version

xrpl-mithril requires Rust 1.85.0 or later (the first stable release supporting edition 2024). MSRV is enforced in CI and will not be raised without a minor version bump.

License

Licensed under either of Apache License, Version 2.0 or MIT license at your option.

Contributing

Contributions are welcome. See CONTRIBUTING.md for guidelines.