Using Skills

Authentication

Before using any skill, you must authenticate:

npx starkfi@latest auth login user@example.com
npx starkfi@latest auth verify user@example.com 123456

The authenticate-wallet skill handles this automatically — just tell your AI agent "log me in" and provide the OTP from your email.

Transaction Skills

Trade (Single Swap)

Swap one token for another using Fibrous aggregation for the best route:

npx starkfi@latest trade 0.1 ETH USDC
npx starkfi@latest trade 100 USDC STRK --slippage 0.5
npx starkfi@latest trade 0.5 ETH DAI --simulate

The --simulate flag estimates fees without executing — useful for previewing before committing.

Multi-Swap

Execute up to 3 swap pairs in a single Starknet transaction:

npx starkfi@latest multi-swap "100 USDC>ETH, 50 DAI>STRK"
npx starkfi@latest multi-swap "0.1 ETH>USDC, 200 USDT>DAI" --simulate

Batch (Multicall)

Combine different operation types — swaps, staking, lending, and sends — in one atomic transaction:

npx starkfi@latest batch \
  --swap "0.1 ETH USDC" \
  --stake "50 STRK karnot" \
  --simulate

Requires at least 2 operations. Each flag (--swap, --stake, --supply, --send) can be repeated.

Staking

Stake STRK, WBTC, tBTC, SolvBTC, or LBTC with a validator:

npx starkfi@latest validators                       # List validators
npx starkfi@latest pools karnot                      # List pools for a validator
npx starkfi@latest stake 100 -v karnot               # Stake (default token: STRK)
npx starkfi@latest rewards --claim -v karnot          # Claim rewards
npx starkfi@latest unstake intent -v karnot -a 50     # Start unstaking
npx starkfi@latest unstake exit -v karnot             # Complete unstaking
npx starkfi@latest stake-status                       # View staking overview

Lending (Vesu V2)

Supply, borrow, repay, and manage lending positions:

npx starkfi@latest lend-pools                           # List pools
npx starkfi@latest lend-supply 100 -p Prime -t USDC     # Supply
npx starkfi@latest lend-borrow -p Prime \                # Borrow
  --collateral-amount 0.5 --collateral-token ETH \
  --borrow-amount 500 --borrow-token USDC
npx starkfi@latest lend-repay 500 -p Prime -t USDC \     # Repay
  --collateral-token ETH
npx starkfi@latest lend-withdraw 100 -p Prime -t USDC   # Withdraw
npx starkfi@latest lend-status                              # View all positions
npx starkfi@latest lend-status -p Prime \                    # View specific position
  --collateral-token ETH

Send

Transfer tokens to another Starknet address:

npx starkfi@latest send 10 STRK 0x04a3...abc
npx starkfi@latest send 0.5 ETH 0x07b1...def --simulate

Data Skills

Balance

Check individual or all token balances:

npx starkfi@latest balance               # All tokens
npx starkfi@latest balance --token STRK   # Specific token
npx starkfi@latest balance --json         # Machine-readable

Portfolio

View a comprehensive DeFi dashboard (balances + staking + lending with USD values):

npx starkfi@latest portfolio
npx starkfi@latest portfolio --json

Utility Skills

Config

Manage RPC, network, gas settings, and transaction status:

npx starkfi@latest config set-rpc https://starknet-mainnet.g.alchemy.com/v2/<key>
npx starkfi@latest config set-network mainnet
npx starkfi@latest config set-gasfree on       # Developer sponsors gas
npx starkfi@latest config set-gas-token USDC    # Pay gas in USDC
npx starkfi@latest config list                  # Show all settings
npx starkfi@latest tx-status 0x0abc...          # Check any transaction

Simulation

Every transaction skill supports --simulate for dry-run execution:

npx starkfi@latest trade 100 USDC ETH --simulate
npx starkfi@latest send 10 STRK 0x04a3... --simulate
npx starkfi@latest batch --swap "0.1 ETH USDC" --stake "50 STRK karnot" --simulate

Simulation returns estimated fees and validates the transaction without spending gas.

JSON Output

Every skill supports --json for machine-readable output:

npx starkfi@latest balance --json
npx starkfi@latest portfolio --json
npx starkfi@latest trade 100 USDC ETH --json

This is especially useful when AI agents need to parse results programmatically.

On this page