Documentation
Kanari DocsDeveloper documentation
Kanari Documentation
Welcome to Kanari Docs
Documentation for the current Kanari Network architecture.
Last updated
Kanari Network
Kanari Network is a Move-based object blockchain with a Mysticeti-style DAG metadata layer, transaction-driven checkpoints, RocksDB-backed state, and JSON-RPC surfaces for wallets, explorers, and SDKs.
The current system is optimized for deterministic execution and long-running validator operations:
- Checkpoint height advances only when committed transaction work exists.
- Empty network DAG vertices can be retained and synchronized without minting empty checkpoints.
- Move execution tracks object/resource state, gas effects, dynamic fields, and deterministic state roots.
- Nodes can recover with strict snapshots or encrypted full-validator backups.
- RPC includes bounded heavy-VM execution and per-client protection, while public deployments should still use a real gateway or load balancer.
Key Features
- Move VM execution: Object, resource, coin, NFT, dynamic-field, and module operations run through the Kanari Move runtime.
- Mysticeti DAG metadata: Validators exchange DAG vertices and checkpoint evidence through libp2p.
- Transaction-driven checkpoints: No transaction means no new checkpoint height, which avoids idle block production.
- Verifiable state: RocksDB state is anchored by an SMT root and checkpoint metadata.
- Operational recovery: Snapshot export/import and encrypted validator backup/restore cover node recovery paths.
- Modern RPC surface: Axum JSON-RPC serves explorer, CLI, SDK, and gateway deployments.
Quick Links
Learn how to build and run the Kanari CLI.
Open resource ArchitectureSee the current layered protocol stack and data flow.
Open resource Multi-node OperationsRun a local 4-validator devnet and understand checkpoint sync.
Open resource Coin TransferUnderstand how to send and receive tokens.
Open resource API ReferenceIntegrate Kanari Network into your applications.
Open resource Security ModelReview key handling, RPC exposure, backup, and test coverage.
Open resourceCurrent mental model
One transaction passes through the RPC layer, enters the verified mempool, is selected into checkpoint work, executes in the Move runtime, updates persistent state, records DAG/checkpoint metadata, and becomes queryable through RPC and explorer indexes.
For operators, the most important invariant is simple:
Checkpoint height means committed chain work, not elapsed time.
That is why idle validators should stop producing new checkpoints until real transactions arrive.