Walrus Docs

Walrus is a verifiable data platform for high-stakes systems that require provable, programmable, always-available data with no performance tradeoffs.

Modern financial systems and AI agents increasingly depend on fast, reliable, and verifiable data. Traditional storage assumes integrity and pushes trust outside the data layer. Walrus embeds availability, integrity, and programmability directly into storage itself.

Walrus enables builders to store large, unstructured data blobs in a way that is:

• Highly available
• Cryptographically verifiable
• Programmable through smart contracts

Core capabilities​

Storage and retrieval​

Walrus supports writing and reading large blobs of unstructured data. Data is content-addressed. Any change to the data produces a new identifier. This makes integrity tamper-evident and enables independent verification of stored content. Walrus also enables anyone to prove that a blob has been stored and remains available for retrieval.

Data availability and fault tolerance​

Walrus uses erasure coding and high redundancy (~4.5x) to maintain availability even under partial node failure.

• Reads remain available with up to 2/3 responsive nodes.
• Writes tolerate up to 1/3 unavailable nodes.

This model is more robust than partial-replication systems and more cost-efficient than full replication.

Cost efficiency​

Through erasure coding, Walrus maintains storage overhead at approximately 5x the size of stored data while delivering strong durability and Byzantine fault tolerance. This enables production-grade availability without full replication costs.

Integration with Sui​

Walrus leverages Sui for coordination, attesting availability, and payments. Storage space is represented as a resource on Sui, which can be owned, split, merged, and transferred. Stored blobs are also represented by objects on Sui, which means that smart contracts can check whether a blob is available and for how long, extend its lifetime, or optionally delete it.

Epochs and WAL​

Walrus is operated by a committee of storage nodes that evolve between epochs. A native token, WAL (and its subdivision FROST, where 1 WAL is equal to 1 billion FROST), is used to delegate stake to storage nodes, and those with high stake become part of the epoch committee. The WAL token is also used for payments for storage. At the end of each epoch, rewards for selecting storage nodes, storing, and serving blobs are distributed to storage nodes and those that stake with them. All these processes are mediated by smart contracts on the Sui platform.

Flexible access​

You can interact with Walrus through a command-line interface (CLI), software development kits (SDKs), and Web2 HTTP technologies. Walrus is designed to work well with traditional caches and content distribution networks (CDNs), while ensuring all operations can also be run using local tools to maximize decentralization.

When to use Walrus​

Use Walrus when your application depends on data that must be:

1. Independently verifiable​

You need to prove where data came from, confirm it has not been altered, or anchor workflows to specific dataset versions.

Examples:

• AI model artifacts and agent memory
• Execution logs for exchanges or markets
• On-chain governance data
• Audit trails for financial systems

2. Highly available under failure​

Your system cannot tolerate downtime, partial node failure, or data loss.

Examples:

• Market infrastructure
• Autonomous agents coordinating state
• Financial protocols with real economic risk

3. Programmable at the data layer​

You need smart contracts to manage, verify, or automate around stored data.

Examples:

• Versioned datasets referenced in AI workflows
• Smart contract-controlled storage lifetimes
• On-chain verification of off-chain artifacts

4. Cost-efficient at scale​

You require strong durability and Byzantine fault tolerance without full-replication overhead.

When not to use Walrus​

Walrus is not optimized for:

• Small, ephemeral application state better suited for direct on-chain storage
• Ultra-low-latency in-memory databases
• Pure archival storage without verification requirements

Walrus is designed for high-stakes systems where availability, integrity, and programmability are structural requirements, not optional features.