Agent Multisig Coordination API is designed with security as a primary concern.

1. Never store private keys - Private keys remain with agents. We only see public keys.
2. Never generate keys for agents - Agents bring their own keys.
3. Never sign transactions - Agents sign; we coordinate.
4. Never hold funds - Multisig addresses are controlled by agents, not us.

┌─────────────────────────────────────────────────────────────────┐
│                        AGENT LAYER                               │
│  Agents keep private keys. Sign digests locally. Submit sigs.   │
└───────────────────────────────┬─────────────────────────────────┘
                                │
┌───────────────────────────────▼─────────────────────────────────┐
│                     COORDINATION LAYER (this API)                │
│  - Creates unsigned transactions (PSBTs)                         │
│  - Computes sighashes for agents to sign                         │
│  - Collects signatures                                           │
│  - Validates signatures match registered pubkeys                 │
│  - Finalizes transactions when threshold met                     │
│  - Broadcasts to network                                         │
└───────────────────────────────┬─────────────────────────────────┘
                                │
┌───────────────────────────────▼─────────────────────────────────┐
│                       BLOCKCHAIN LAYER                           │
│  Bitcoin, EVM, Stacks - native multisig verification            │
└─────────────────────────────────────────────────────────────────┘

We use PSBTs (BIP-174) for Bitcoin transactions:

1. Full visibility - Agents can decode PSBT to see all inputs/outputs
2. Partial signing - Each agent signs only their part
3. Validation - Agents should verify:
   • Input amounts match expected
   • Output addresses are correct
   • Fee is reasonable

Our Bitcoin multisigs use Taproot (P2TR) with script-path spending:

• Privacy - Script not revealed until spending
• Schnorr signatures - BIP-340 compliant
• No key aggregation - Each key signs independently (safer)

For EVM chains, we use Safe (formerly Gnosis Safe):

• Battle-tested - Secures $100B+ in assets
• EIP-712 signatures - Human-readable signing
• Deterministic addresses - CREATE2 deployment
• No proxy pattern - Direct contract interaction

DO NOT open public issues for security vulnerabilities.

Email: security@agentmultisig.dev (not yet active)

Or DM @AetosSET on Twitter.

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

• 24 hours - Acknowledgment
• 72 hours - Initial assessment
• 7 days - Fix deployed (if critical)
• 30 days - Public disclosure (coordinated)

Before going to production:

• Verify you're connecting to the correct API endpoint
• Validate sighashes before signing
• Check output addresses match expected destinations
• Review fee amounts are reasonable
• Test with small amounts first
• Monitor your multisig addresses for unexpected activity
• Keep agent private keys secure (HSM recommended for production)
• Implement webhook signature verification (coming soon)
• Set up alerts for proposal creation

Not yet audited.

This is pre-production software. Use at your own risk.

We plan to pursue a security audit before v1.0 release.

1. In-memory storage - Default mode doesn't persist across restarts
2. No HSM support - Agents must manage their own key security
3. Rate limiting - Basic implementation, not production-hardened
4. No MFA - API keys are single-factor

No formal bug bounty program yet.

Significant vulnerabilities may be rewarded at our discretion.