Two jobs. Do not mix them.

A local kaspad on Testnet-10 plus a CPU miner paying a kaspatest: address. tKAS is faucet money. Official line: CPU only so you do not own a tiny testnet.

GitHub puts the zip in Downloads. Do not run from there. Browser plus OneDrive plus kaspa-miner (3).exe is how paths break.

Get rusty-kaspa latest ↗: Windows rusty-kaspa-*-win64.zip, Linux rusty-kaspa-*-linux-amd64.zip, macOS rusty-kaspa-*-osx.zip. CPU miner: kaspanet/cpuminer ↗.

Make a dedicated folder and copy the unzipped files there. Windows: C:\Kaspa\tn10\. Ubuntu / Linux / macOS: ~/kaspa/tn10/. Confirm you see kaspad.exe or kaspad (sometimes in bin/).

Windows after unzip:

New-Item -ItemType Directory -Force -Path C:\Kaspa\tn10
Copy-Item "$env:USERPROFILE\Downloads\rusty-kaspa-*\*" C:\Kaspa\tn10 -Recurse

Linux / macOS:

mkdir -p ~/kaspa/tn10
unzip -o ~/Downloads/rusty-kaspa-*-linux-amd64.zip -d ~/kaspa/tn10
chmod +x ~/kaspa/tn10/kaspad ~/kaspa/tn10/kaspa-wallet ~/kaspa/tn10/rothschild

If the zip dumped a bin/ subfolder, cd there. Copy the miner into the same folder. Rename long GitHub names: kaspa-miner-v0.2.7-win64-amd64 (3).exe becomes kaspa-miner.exe.

Must start with kaspatest:. A kaspa: address is the wrong network.

Kaspa-NG ↗: network Testnet-10, create wallet, copy receive address. CLI: kaspa-wallet in the node zip, against a running TN10 node with --utxoindex. Throwaway: rothschild in the same zip can print a key plus address. Save the key if you want that bag.

Paste the full string into tn10.kaspa.stream ↗ search. If it rejects it, you chopped the address or used mainnet. Faucet is optional: faucet-tn10.kaspanet.io ↗. You do not need it to mine.

Windows PowerShell:

cd C:\Kaspa\tn10
dir kaspad.exe
.\kaspad.exe --testnet --netsuffix=10 --utxoindex

Ubuntu / Linux / macOS:

cd ~/kaspa/tn10
./kaspad --testnet --netsuffix=10 --utxoindex

One line. Enter once. Closing this window kills the node. One kaspad only. Windows Firewall: allow private networks.

Fails already hit: command from your home folder is “not recognized” (you are not in the dedicated folder). Two commands glued on one line makes --rpclisten-borsh=default.\kaspad.exe and “invalid IP”. Missing --testnet puts you on mainnet by accident.

Order: validating pruning-point proof (levels counting down). Receiving UTXO set chunks (--utxoindex is slow, hundreds of millions of UTXOs). “Finished receiving the UTXO set” then importing (looks frozen; do not kill it). Live tip: Accepted N blocks via relay every second.

Only step 4 is ready. Plan 1 to 2 hours on a home line.

Node stays up. You do not hard-close kaspad to free the miner. Rename the long miner filename first if needed.

Windows:

cd C:\Kaspa\tn10
.\kaspa-miner.exe --testnet --mining-address kaspatest:YOUR_ADDRESS -s 127.0.0.1 -p 16210 -t 1

Linux / macOS:

cd ~/kaspa/tn10
./kaspa-miner --testnet --mining-address kaspatest:YOUR_ADDRESS -s 127.0.0.1 -p 16210 -t 1

Port 16210 is TN10 gRPC, not mainnet 16110. -t 1 is one CPU thread. Enough on TN10.

tn10.kaspa.stream address lookup ↗. explorer-tn10.kaspa.org was unreliable. Coinbase stays Confirming until mature. Blue good, red bad. Then bring the address to the playground.

Change address: Ctrl+C the miner only. Restart with the new kaspatest: string. Old coins stay on the old address.

Official line is still CPU only on TN10. GPU is allowed by physics, not by etiquette. One card can own a network that sits around tens of MH/s. There is no public GPU pool for tKAS.

BzMiner to the node: bzminer -a kaspa -w kaspatest:YOUR_ADDRESS -p node+tcp://127.0.0.1:16210 --nc 1. Some builds want solo+tcp://. Do not use stratum+tcp://pool....

Or tmrlvi/kaspa-miner ↗ GPU build, -t 0 for GPU only. After Toccata, old miners can hash and still submit invalid blocks. Then use stratum-bridge.exe from the same zip, miner to 127.0.0.1:5555. Do not port-forward stratum.

Follow real KAS. This is not a miner guide. Do not add --testnet. Same release zip, different folder: C:\Kaspa\mainnet\ or ~/kaspa/mainnet/.

SSD. 16 GB RAM minimum, 32 GB more comfortable. 4+ cores. Tens of GB free. Stable link. Mainnet after Crescendo and Toccata is heavier than TN10.

Linux / Ubuntu / macOS:

cd ~/kaspa/mainnet
./kaspad --utxoindex

Windows:

cd C:\Kaspa\mainnet
.\kaspad.exe --utxoindex

Optional local RPC. Keep it local:

./kaspad --utxoindex --rpclisten=127.0.0.1:16110

Do not expose 16110. IBD is long. Same rule: do not kill during UTXO import. One process per data dir.

Logs look like live tip: accepted blocks via relay, steady processed N blocks, not receiving UTXO set or pruning-level validation. Then a wallet can attach to 127.0.0.1:16110.

Foreground is fine for learning. Linux user systemd or tmux on Mac. One kaspad per data dir. Two processes on the same data dir is a lock or corruption.

Update: stop cleanly, overwrite kaspad, start again. Stay on 2.0.1+ (Toccata). Same data dir. Do not delete the DB unless the release notes say to resync.

You can sync forever as a private follower. Public means others can dial in on P2P 16111/tcp on a real public IP.

Router: forward 16111 to this machine. Firewall allow 16111 (sudo ufw allow 16111/tcp on Ubuntu). Do not bind P2P to 127.0.0.1. Need a real public IP. CGNAT means you will never list. RPC stays private. Public node is not open RPC.

kaspa.stream/nodes ↗. Use Check your node. If the crawler sees you, the page highlights you. Want user agent kaspad:2.0.1 (Toccata). Wait 10 to 30 minutes after opening the port.

Extra port test: kaspa.host ↗ on mainnet 16111. Not listed means private follower. That is still a valid node. The chain does not require you to be on the public list.