Purpose: To provide AI agents with self-custodial DeFi operation capabilities, covering Ethereum, Arbitrum, Polygon, Base, Optimism, TRON, Solana, Bitcoin, and Litecoin. The design premise is that the agent, MCP server, and host may all be compromised, and only the Ledger device is trusted; therefore, every transaction can be verified on the device screen, and tampering (such as replacing the recipient address, rewriting the swap path, or inserting an approval) can be detected.
Main capabilities: Read cross-chain balances, DeFi positions, NFT portfolios, profit and loss, and daily briefings; support Aave V3, Compound V3, Morpho Blue, Uniswap V3, Curve, Lido, EigenLayer, Rocket Pool, Safe multisig, MarginFi, Kamino, Marinade, Jito, SunSwap, as well as LiFi cross-chain swaps/bridging and Jupiter v6; provide about 190 tools including staking, swapping, lending, LP, BTC/LTC sending and PSBT multisig, contract security and risk scoring, transaction decoding, and secondary LLM cross-validation.
Usage: Requires Node.js >= 18.17, and connects via stdio to MCP clients such as Claude Code and Cursor. EVM signs through WalletConnect to Ledger Live; TRON, Solana, BTC, and LTC connect directly to the device via USB HID. For EVM, each transaction requires preview_send before send_transaction; for Solana, preview_solana_send is required first. Optional installation of companion security skills and /setup guidance.
Limitations: The Claude.ai web/desktop version requires a hosted MCP endpoint, which is not yet provided; Ledger Live's WalletConnect bridge does not support the tron, Solana, or BTC/LTC namespaces; some protocols are limited to specific chains; some Solana operations require enabling blind signing; public RPCs have rate limits, and for actual use it is recommended to provide your own RPC.
