The agent you trust.Unseen by everyone else.
Arsene is the trust, privacy and policy layer for autonomous AI agents on Solana. Verifiable identity. Private x402 payment rails. On-chain guardrails — through a single SDK.
Agents are transacting at venture scale on infrastructure built for humans.
Fragmented trust
No standard way to verify that a wallet signing an x402 payment is actually the agent it claims to be. When an agent is hijacked or prompt-injected, there is no protocol-level way to revoke it.
Privacy collapse
Every agent action is a glass box. Spending patterns, vendor relationships, operating cadence — all public. For enterprise agents handling treasury or procurement, this is a non-starter.
Policy vacuum
Spending limits live in application code. One jailbreak, one RPC swap, one malicious MCP server — and the guardrails evaporate. Nothing is enforced at the protocol level.
A mask, a shadow, and a lock.
Three primitives. One SDK. Four lines in ElizaOS, LangGraph or the Vercel AI SDK and your agent inherits the full kit.
Masque
Every agent wears a Masque — a sovereign cryptographic identity anchored on Solana. Principals issue signed credentials defining scope, lifespan, revocation keys. Compromised masks burn on-chain, and every merchant in the network rejects that agent on the next block.
Ombre
The shadow your agent casts. Ombre wraps x402 with selective disclosure — stealth addresses and compressed accounts produce zero-knowledge proofs that payments satisfy policy without publishing sender, receiver, or amount to the public graph.
Serrure
The lock. Serrure is an on-chain policy engine enforcing spending limits, rate limits, allowlists, and circuit breakers at the program level. A compromised agent literally cannot exceed its principal’s rules — the instruction fails before settlement.