Project

Amaru

An alternative Rust node client for the Cardano blockchain. Aimed as an entrypoint for client applications initially, it aspires to become a full block-producing node running side-by-side with the Haskell nodes to further increase the decentralisation of the platform while fostering its rich open source ecosystem.

Goal

Build a new fully interoperable block-producing Cardano node.

One of the core objectives is to achieve a full block-producing Cardano node which can run side-by-side the existing Haskell nodes in a fully interoperable manner. The project is ambitious and spans over many areas such as networking, cryptography or language theory. It is developed from several parts each covering a different aspect of the node and its surrounding ecosystem.

Besides, Amaru is also geared towards developer experience and aims at providing easy to grasp interfaces using widely spread technologies such as gRPC while maintaining compatibility with the strong ecosystem of middlewares, such as Ogmios or Oura, that constitutes the backbone of Cardano services.

Maintainers
Journey

Data node

A node capable of synchronizing the chain from a (remote) trusted peer, and serve the data to client applications through an API query layer.

Such a node is most useful for decentralised applications building on the blockchain. It doesn't however take part in the consensus nor help to propagate blocks on the network.

A full relay node as an entrypoint to the network is still required at this point. This first phase is fully demonstrated by Dolos: a data-node written in Rust and whose core modules will be re-used for the development of Amaru.

Client node

A node capable of synchronizing the chain from a remote trusted peer while maintaining its own state and performing some ledger validations. So while it can only follow a single chain (provided by its single peer), it is fully autonomous in validating that this chain is conform to the protocol.

It can, in particular, validate a slot-leader schedule and check for consensus rewards in full autonomy. Such a node is most useful as a step towards a relay. It doesn't however take part in the consensus nor help to propagate blocks on the network yet.

Relay node

A node capable of validating blocks seen on the network and propagate them by taking part in the p2p gossiping between nodes. It can also seemingly follow the chain from multiple peers by performing adequate chain selection.

Such node doesn't yet take part in the consensus and cannot produce blocks. It can however fully replace any relay. From the perspective of any external observer, it is a Cardano relay node.

Block producing node

A node that can produce blocks and take part to the consensus. It can monitor the leader schedule, manage a mempool, forge blocks and follow protocol updates (hard-forks).

A full node is not however capable of validating the historical chain. Instead, it always bootstrap from a snapshot constructed from an archive node.

Soft-forks & performance improvements

At this point our node should be able to do everything the Haskell node does but with less resources usage and with better performance on most aspects of execution.

Demos

Glimpses of development

Ledger rules, time travelling & node-to-node conversations.

June 27, 2025

A demo showcasing the latest progress on the ledger rules and the Rust-agnostic testing approach used for conformance. Then, we dive into some preliminary visualisation of the simulation engine, with auto-generated sequence diagrams and a terminal user interface that allows exploring execution traces.

Finally, another look at some of the Amaru metrics, obtained from running a cluster of Amaru nodes talking to one another. This demonstrates how our observability stack is particularly well suited for aggregating distributed traces from multiple processes.

Deterministic simulations, ledger rules& P2P Networking

March 28, 2025

First, an overview showcasing the latest progress on deterministically simulating the consensus, using pre-generated blockchain trees.

Then, a tour of the ledger rules and the separation of concerns between block validation and state management, and how Amaru tackles this nicely.

Finally, an update on the ongoing work on the Peer-to-Peer (P2P) networking stack being designed not only for Amaru, but for the wider Rust blockchain ecosystem.

Stake distribution, multi-chain consensus & simulation testing.

February 7, 2025

Three contributions covering:

  1. Progresses on the ledger state tracking and the stake distribution reconstruction now complete up to the first governance proposal refund;
  2. A focus on the consensus with a basic chain selection strategy between two chains coming from two distinct peers;
  3. An introduction to our simulation testing approach inspired from Jepsen's Maelstrőm.

On-disk Ledger State & Observability

December 20, 2024

Showcasing Amaru's on-disk ledger storage, solving a long-standing challenge of the Cardano node. The ledger state in this demo is comprised of the entire UTxO, stake pools parameters and registrations, rewards accounts and delegations. It is tracked throughout block application and persisted on-disk efficiently with regular snapshots at each epoch boundary.

While it doesn't yet calculate rewards at the epoch boundary, the on-disk state now contains all elements necessary to the calculation of rewards. It thus becomes the immediate next step.

First Steps

October 20, 2024

A simple pipeline showcasing an Amaru node fetching blocks from the network, validating their header (VRF & KES) and forwarding them to an in-memory ledger performing UTxO management and phase-2 validations.

Getting started

Explore and contribute.

Quarterly reports

Follow along