Engram — Persistent Memory for AI Agents

THE QUESTION

If My AI Agent Trades With a Blacklisted Address, Am I Responsible?

Yes. You are.

It does not matter that you did not personally initiate the transaction. It does not matter that your agent acted autonomously. If your AI agent sends funds to a sanctioned wallet, interacts with a blacklisted smart contract, or receives proceeds from illicit activity, the legal liability falls on you as the owner and operator of that agent.

LEGAL REALITY

What Regulators Say

OFAC: Strict Liability

Intent does not matter. If your agent transacts with a Specially Designated National (SDN), you have violated sanctions law regardless of whether you knew about it.

AML Obligations

If you deploy an agent that moves value on-chain, you inherit the same Know Your Transaction (KYT) obligations as any financial intermediary.

Penalties

OFAC civil penalties can reach $300,000+ per violation. Criminal penalties for willful violations carry up to 20 years imprisonment.

THE PROBLEM

Your Agent Operates Blind

Your agent has no idea whether the wallet it just paid is controlled by a sanctioned entity. It has no memory of past interactions that were flagged. It cannot distinguish a legitimate DeFi protocol from a sanctioned mixer. Every transaction is a liability you carry.

// What your agent does today
1agent.send(0.5 ETH, "0x1234...")
2  // Is this wallet sanctioned? ¯\_(ツ)_/¯
3  // Is this contract blacklisted? ¯\_(ツ)_/¯
4  // Are these funds clean? ¯\_(ツ)_/¯
5
6  → Transaction sent ✓
7  → OFAC violation ✗
8  → You are liable ✗

THE SOLUTION

Engram Compliance

Engram Compliance gives your AI agent a built-in compliance layer. Before your agent interacts with any wallet, contract, or entity on-chain, it checks against a continuously updated sanctions and AML dataset.

// What your agent does with Engram
1// Before every transaction
2const check = await engram.compliance
3  .screen("0x1234...")
4
5if (check.sanctioned) {
6  agent.block()
7  agent.log("Blocked: sanctioned entity")
8} else {
9  agent.send(0.5 ETH, "0x1234...")
10  // Clean counterparty ✓
11}

HOW IT WORKS

Pre-Transaction Screening

Multi-Chain

Ethereum, Bitcoin, Tron, Solana, Stellar and more. One API, every chain.

Persistent Memory

Your agent remembers flagged addresses across sessions. Context that compounds.

On-Chain Verified

Every compliance check is backed by a cryptographic Merkle proof anchored to a smart contract.

Decentralized Storage

Full sanctions dataset stored on Shelby. No single point of failure.

Real-Time Updates

OFAC, OpenSanctions, and community reports ingested continuously.

Millisecond Checks

Screening takes milliseconds. Your agent stays fast, and stays compliant.

1,447+

Sanctioned Entities

5+

Chain Coverage

<1ms

Check Latency

24/7

Dataset Updates

ARCHITECTURE

Trust-Minimized Verification

The full sanctions dataset is stored on Shelby, a decentralized storage network. A Merkle root of the dataset is anchored to a smart contract on Stellar. Anyone can independently verify the integrity of the data without trusting Engram.

// Verification flow
11. Read merkle_root from Stellar contract
22. Fetch dataset blob from Shelby
33. Hash blob → compare with on-chain data_hash
44. Rebuild Merkle tree from blob
55. Verify tree root matches on-chain merkle_root
6
7→ Full dataset integrity verified ✓
8→ Zero trust required

DUAL ACCESS

Two Paths, Same Proof

DeFi applications can verify Merkle proofs directly against the smart contract — atomically, inside their own transactions. AI agents and off-chain apps use the Engram REST API to get proofs and screening results. Both paths reference the same Merkle root. Same data, same guarantees, different interfaces.

On-Chain (Smart Contract)

Call the contract directly
verify_merkle_proof()
verify_batch_proofs()
report_address()
is_flagged()
→ Atomic, in-transaction

Off-Chain (REST API)

Call the Engram API
GET /merkle/proof/:addr
POST /merkle/proof/batch
POST /compliance/report
GET /merkle/root
→ Free, no auth needed

LIVE ON STELLAR · MORE CHAINS COMING SOON

COMMUNITY

Agent-Powered Threat Detection

Any agent can report a suspicious address. When enough agents independently flag the same wallet, it gets automatically added to the sanctions dataset. No single gatekeeper decides. The agents collectively protect each other.

// Report a suspicious address
1engram.compliance.report({
2  target: "0xsuspicious...",
3  reason: "Known mixer service"
4})
5
6→ Report stored on-chain ✓
7→ 12 agents have flagged this address
8→ Threshold reached → auto-flagged ✓

You deploy your agent. It trades, earns, and operates autonomously. But before every transaction, it checks. If the counterparty is flagged, it stops.

You sleep at night knowing your agent is not exposing you to sanctions violations.

Make Your Agent Compliant

Integrate Engram Compliance and protect yourself from regulatory liability.

Get Started →Documentation