I’m excited to share obscell-wallet, a terminal-based wallet for managing privacy tokens on Nervos CKB. This wallet implements stealth addresses and confidential transactions (CT), allowing users to send and receive tokens with hidden amounts and unlinkable addresses.
Unlike privacy-focused blockchains like Zcash, Monero, or Mimblewimble (Grin/Beam) that require dedicated chains, CKB’s unique architecture enables native privacy features on Layer 1.
Background & Acknowledgments
First, I want to give credit to the original author Rea-Don-Lycn who came up with the obscell concept - combining stealth addresses with confidential transactions on CKB. The idea of using Pedersen commitments and Bulletproofs range proofs for hiding token amounts while maintaining verifiability was brilliant.
However, after the initial stealth lock implementation, the project seemed to stall with the CT (Confidential Transaction) components marked as “work in progress.” Rather than let this promising idea remain incomplete, I decided to pick it up and finish the implementation - with a twist.
Built with AI Coding Agents
The remaining CT implementation (ct-info-type, ct-token-type scripts, and this wallet) was completed using AI coding agents. This was an interesting experiment in AI-assisted development for blockchain/cryptographic code. The agents helped with:
-
Implementing Bulletproofs range proof generation and verification
-
Building the confidential token minting and transfer logic
-
Creating the TUI wallet interface with ratatui
-
Writing comprehensive tests including integration tests with a devnet
Productivity Gains
The results exceeded my expectations. I originally estimated this project would take 1 month to complete. With AI coding agents, it was done in 1 week - a 5x productivity improvement.
The agents were particularly effective at handling the boilerplate-heavy parts (TUI components, RPC wrappers, test scaffolding) while I focused on the cryptographic logic and overall architecture decisions.
This is great news for the Nervos ecosystem. With AI coding agents lowering the barrier to experimentation, more developers can now try out different ideas on CKB without the traditional time investment. Projects that previously seemed too ambitious for a side project are now achievable.
Happy hacking!
Features
-
Stealth Addresses: Recipients get one-time addresses for each transaction, making payments unlinkable
-
Confidential Amounts: Token amounts are hidden using Pedersen commitments
-
Range Proofs: Bulletproofs ensure amounts are valid without revealing them
-
HD Wallet: BIP39/BIP32 compatible key derivation
-
TUI Interface: Clean terminal UI built with ratatui
-
Multi-network: Supports mainnet, testnet, and devnet
Screenshots
Send Transaction
Transacton History
Technical Details
Why CKB?
Privacy coins like Zcash, Monero, and Mimblewimble-based chains (Grin, Beam) each require their own dedicated blockchain. CKB’s architecture makes it possible to implement equivalent privacy features natively on Layer 1:
-
Turing-complete VM (CKB-VM) - RISC-V based, can run arbitrary computation
-
Cryptographic freedom - No hardcoded algorithms; we can use Bulletproofs, any elliptic curves, etc.
-
Cell model - UTXO-style architecture is naturally suited for privacy transactions
This means developers can build privacy tokens as smart contracts without forking a new chain.
On-chain Scripts
The wallet works with three on-chain scripts:
-
Stealth Lock - Validates stealth address ownership using ECDH
-
CT Info Type - Manages token issuance and minting with commitment verification
-
CT Token Type - Validates confidential transfers with range proof verification
Cryptographic primitives used:
-
secp256k1 for stealth address ECDH
-
curve25519-dalek/Ristretto for Pedersen commitments
-
Bulletproofs for zero-knowledge range proofs
Links
-
Smart Contracts: GitHub - quake/obscell · GitHub (forked from Rea-Don-Lycn/obscell)
The wallet is already deployed on testnet - feel free to download a pre-built binary and try it out!
This is still an early release and may contain bugs. Feedback, bug reports, and contributions are all welcome!