Build on Kaspa with AI
HardKAS is a local-first development framework, transaction planner/signer, SilverScript toolchain wrapper, and artifact verifier for Kaspa L1.
[!IMPORTANT] Kaspa L1 is a Proof-of-Work UTXO BlockDAG, NOT an EVM.
- Do not write Solidity or EVM RPC scripts for Kaspa L1.
- Do not write custom
kaspa-wasmpayment/signing scripts when a HardKAS CLI command or SDK method already implements the operation.- Do not treat
simulatedledger runs or localsilver test(SCRIPT_VM) runs as real node consensus proof.
Versions & Pinned Toolchain
- Repository development version:
0.13.0-rc.3(may be ahead of npmlatest; checknpm view @hardkas/cli dist-tagsand usenpx @hardkas/cli@latestwhen installing from npm outside the monorepo) - Localnet node (
--profile toccata-v2):rusty-kaspadv2.1.0(containerhardkas-kaspad-toccata-v2, JSON wRPC port18210; custom Docker setups may publish another port such as26510) - SilverScript compiler:
silvercv1.0.0(installed viahardkas toolchain install silverc) - Node.js requirement:
24.22.1
What HardKAS Does (and Does Not Do)
| Area | Status in this release | Notes for agents |
|---|---|---|
Payments (tx plan, tx sign, tx send, tx wait, tx status) | Verified on simnet | Uses upstream kaspa-wasm Generator; supports Schnorr & ECDSA, CLTV (--lock-time-daa / --lock-time-ms), and multi-recipient batches. |
SilverScript P2SH (silver compile, verify, deploy, spend) | Verified on simnet | Compiles .sil with pinned silverc v1.0.0, deploys and spends v0 P2SH contracts. |
Toccata 1:1 Covenants (silver covenant genesis, transition, inspect) | Verified on simnet | Real v1 transactions on rusty-kaspad v2.1.0 with covenant_id lineage checks. |
Artifact & Receipt Verification (artifact inspect, artifact verify, tx receipt) | Verified | Integrity check works offline on all networks; network & consensus checks query a real node of the artifact's network. |
silver covenant start|advance (1:N, N:M), kcc20, silver test --on-node, zk toccata spend, testnet-10 | Experimental | Explicitly opt-in or experimental; do not present them as stable. |
silver test, silver debug, silver preflight (local runner) | Unverified (SCRIPT_VM) | Requires --allow-unverified-runner with a local build; never independent consensus proof. Use --on-node or real node spends for consensus checks. |
mainnet (reads, signing, broadcast) | Blocked by policy | Refused before any network connection (MAINNET_CONNECTION_DISABLED, MAINNET_SIGNING_DISABLED, MAINNET_BROADCAST_DISABLED). |
vProgs, trustless bridges, @hardkas/l2, @hardkas/pskt-native | Not supported / Not published | Only offline artifact inspection for vProgs; do not invent bridge or L2 commands. |
Environment Matrix
| Environment | network flag | Node required | Consensus proof? | Typical use |
|---|---|---|---|---|
| In-memory ledger | simulated | No | No (synthetic UTXOs, no script VM) | Fast local loop, replay verify |
| Managed or custom simnet | simnet | Yes (rusty-kaspad 2.1.0) | Yes (local node consensus) | Payments, SilverScript P2SH, Toccata v1 covenants |
| Public Testnet-10 | testnet-10 | Yes (opt-in) | Yes (experimental) | Public test funds only (experimental: true, allowPublic: true) |
| Mainnet | mainnet | Refused | Offline artifact verify only | Blocked for RPC, signing, and broadcast |
First Checks Before Running Tasks
Run these machine-readable checks first so you never assume unavailable features:
hardkas doctor --json
hardkas silver doctor --network simnet --json
In TypeScript SDK code, call sdk.capabilities.get() before invoking optional or experimental namespaces.
Canonical Entry Points
- Step-by-step agent recipes (6 core tasks): Canonical Workflows
- Error codes & recovery: AI Troubleshooting
- Agent setup for Claude Code & Cursor: Claude Code & Cursor Guide
- Machine-readable maps:
/llms.txt(short navigation map) and/llms-full.txt(complete AI reference) - Full human guide for simnet:
docs/start/build-on-simnet.mdin the repository - Video tutorials: KasLabDevs on YouTube