Skip to main content

BlockTrails

Provably safe agents via Bitcoin state anchoring. Full audit trail. Trust through time.

The Problem​

How do you trust an AI agent? Traditional approaches rely on:

  • Reputation — But reputation can be gamed
  • Certification — But who certifies the certifiers?
  • Promises — But promises can be broken

BlockTrails offers a different approach: trust through time. An agent that has behaved safely for months, with cryptographic proof, is more trustworthy than one making promises.

How It Works​

Agent behavior is cryptographically anchored to Bitcoin:

┌──────────────────────────────────────────────────────────┐
│                     Agent Action Log                      │
├──────────────────────────────────────────────────────────┤
│  Timestamp: 2025-01-19T12:00:00Z                         │
│  Agent: did:nostr:npub1abc123...                         │
│  Action: read_file                                        │
│  Resource: https://alice.pod/documents/notes.ttl         │
│  Result: success                                          │
│  Hash: sha256:abc123...                                   │
└──────────────────────────────────────────────────────────┘
                           │
                           ▼
              ┌────────────────────────┐
              │  Merkle Tree Batch     │
              │  (many action hashes)  │
              └────────────────────────┘
                           │
                           ▼
              ┌────────────────────────┐
              │  Bitcoin Transaction   │
              │  OP_RETURN: root hash  │
              └────────────────────────┘
  1. Agent takes action — Read, write, API call, etc.
  2. Action is logged — Structured record with timestamp
  3. Actions are batched — Collected into Merkle tree
  4. Root hash anchored — Single Bitcoin OP_RETURN per batch
  5. Proof generated — Merkle path proves inclusion

Verification​

Anyone can verify an agent's history:

import { verifyTrail } from 'blocktrails';

const verification = await verifyTrail({
  agentDid: 'did:nostr:npub1abc123...',
  action: actionLog,
  merkleProof: proof,
  bitcoinTx: txid
});

if (verification.valid) {
  console.log('Action proven at:', verification.timestamp);
  console.log('Bitcoin block:', verification.blockHeight);
}

Why Bitcoin?​

  • Immutable timestamping — Can't rewrite history
  • Decentralized verification — No trusted third party
  • Long-term availability — Bitcoin isn't going away
  • Cost efficiency — Batch thousands of actions per transaction
  • Global consensus — Everyone agrees on the timeline

Trust Score​

Over time, agents build verifiable track records:

MetricDescription
AgeHow long the agent has been operating
VolumeNumber of verified actions
ConsistencyRegularity of anchoring
ScopeTypes of actions performed
IncidentsAny flagged behaviors

Integration with SAND​

BlockTrails integrates with the SAND stack:

  • Solid — Log access to pod resources
  • ActivityPub — Track federation activities
  • Nostr — Verify event publishing history
  • DID — Agent identity via did:nostr

Use Cases​

AI Assistant Verification​

Prove your AI assistant only accessed files you permitted, when you permitted.

Compliance Auditing​

Demonstrate regulatory compliance with cryptographic proof.

Multi-Agent Systems​

Agents can verify each other's track records before collaboration.

Data Stewardship​

Prove chain of custody for sensitive data handling.

See Also​