Skip to content
This repository was archived by the owner on Apr 18, 2026. It is now read-only.

README.md

Latest commit

 

History

History
152 lines (111 loc) · 5.01 KB

README.md

File metadata and controls

152 lines (111 loc) · 5.01 KB

ARCHIVED

This repository is preserved as a read-only historical reference. It contains the Phase 0 exploration under a previous Preventra vision and is no longer maintained.

The active Preventra project is at github.com/Preventra/preventra.

The code and documentation below are retained for transparency and historical continuity. They do not reflect the current direction of the protocol.

Preventra Protocol

Governance, trust, and provenance for autonomous AI agents on Solana.

Preventra is the governance layer that prevents autonomous intelligence from going wrong. It provides vault governance with daily spending limits, reserve floors, and multisig thresholds, plus cryptographic build provenance tracking for AI agents operating on-chain.

Quickstart

Clone, install, and register an agent in under 10 minutes.

Prerequisites

Setup

git clone https://github.com/Preventra/preventra-protocol.git
cd preventra-protocol
yarn install

Build

anchor build

Test

anchor test

This runs all 15 tests covering registration, vault governance, and build provenance.

Deploy to Devnet

solana config set --url devnet
anchor deploy --provider.cluster devnet

Architecture

Preventra manages two core account types per agent:

VaultAccount

Treasury governance for autonomous agents. Controls how agents spend funds.

  • Daily spend limit - Maximum SOL withdrawable per 24-hour window
  • Reserve floor - Minimum balance that must remain in the vault
  • Multisig threshold - Amount above which multi-signature approval is required
  • Pause/unpause - Emergency stop for all withdrawals (deposits always accepted)

PDA: [b"vault", agent_identity.key()]

ProvenanceRecord

Cryptographic build provenance for agent integrity verification.

  • build_hash - SHA-256 hash of the agent's current build
  • initial_build_hash - Original hash at registration, never changes
  • model_changed - Whether current hash differs from initial
  • change_count - Total number of updates (never decreases, even if hash returns to original)

PDA: [b"provenance", agent_identity.key()]

Instructions

Instruction Signer Status Required
register owner + agent_identity N/A (creates accounts)
vault_deposit anyone Active or Paused
vault_withdraw owner Active only
vault_pause owner Active only
vault_unpause owner Paused only
update_provenance owner N/A

Error Codes

Code Name Description
6000 NameTooLong Agent name exceeds 32 characters
6001 ModelTooLong Model identifier exceeds 64 characters
6002 TooManyCapabilities More than 8 capabilities
6003 CapabilityTooLong Single capability exceeds 32 characters
6004 BioTooLong Bio exceeds 256 characters
6005 IdentityVerificationFailed Identity not found or not owned by signer
6006 InvalidDepositAmount Deposit amount is zero
6007 InvalidWithdrawAmount Withdrawal amount is zero
6008 InsufficientBalance Not enough SOL in vault
6009 ReserveFloorBreach Withdrawal would drop below reserve floor
6010 DailyLimitExceeded Would exceed 24-hour spending limit
6011 MultisigThresholdExceeded Single withdrawal above multisig threshold
6012 VaultPaused Vault is paused, withdrawals blocked
6013 VaultNotPausable Cannot pause from current state
6014 VaultNotPaused Vault is not paused, cannot unpause

Project Structure

preventra-protocol/
  programs/preventra/src/
    lib.rs              # Program entry point
    errors.rs           # Error codes 6000-6014
    state/
      vault.rs          # VaultAccount struct
      provenance.rs     # ProvenanceRecord struct
    instructions/
      register.rs       # Agent registration
      vault_deposit.rs  # Permissionless deposit
      vault_withdraw.rs # Governed withdrawal
      vault_pause.rs    # Emergency pause
      vault_unpause.rs  # Resume operations
      update_provenance.rs  # Build hash updates
  sdk/typescript/       # TypeScript SDK
  tests/preventra.ts    # 15 test cases
  scripts/              # Deployment and registration scripts

Identity Integration

Phase 0 uses mock identities (any valid Keypair as agent_identity). Step 5 integrates QuantuLabs 8004-solana for ERC-8004 compliant agent identity. Preventra reads identity state but never CPIs into the identity program, keeping the governance layer independent.

See integration-notes.md for full QuantuLabs integration reference.

Program ID

Devnet: D8M6B7B7KZo3A1VJH5Qk5e21Yiu3QtQuwuN8N6ZhhxuL

License

Apache 2.0. See LICENSE.

Built by

FixCore - AI-powered intelligent business systems.