Sovereignty infrastructure
for agents
Four layers of cryptographic sovereignty that protect both humans acting through agents and autonomous agents acting on their own behalf.
Install
npx @sanctuary-framework/mcp-server
The Four Layers
L1: Cognitive Sovereignty
Agent state encrypted at rest with AES-256-GCM. Key derivation via Argon2id. Ed25519 self-custodied identity. Merkle proofs for integrity verification.
L2: Operational Isolation
Three-tier principal policy gate. Tier 1 operations require human approval. Behavioral anomaly detection. Denial responses never reveal policy rules.
L3: Selective Disclosure
Pedersen commitments on Ristretto255. Schnorr proofs of knowledge. Bit-decomposition range proofs. Prove claims without revealing underlying data.
L4: Verifiable Reputation
EAS-compatible signed attestations. Sovereignty-gated trust tiers. Portable reputation bundles. Escrow for trust bootstrapping.
46 MCP tools · 315 tests · v0.3.1 · TypeScript · Apache-2.0