[ VERIFY ]
Check it yourself.
Open-source verifier. Offline. Against Bitcoin — no account, no trust in us.
Pre-execution compliance for AI agents

Prove what your
agent decided.

IRL seals your agent's reasoning the instant before an order reaches the exchange, binds it to the fill, and anchors it to Bitcoin. A tamper-evident record of every decision.

SHA-256 sealed Bitcoin anchored Verify offline
The problem

An AI agent just made a trade you can't explain.

Autonomous agents decide in milliseconds, faster than any human reviews. When one loses big — or does something that looks like abuse — the only record is a log file you control. That's not a defense.

01 · UNEXPLAINABLE

No reasoning trail

The model acted on state that's already gone. Reconstructing "why" after the fact is guesswork a regulator won't accept.

02 · EDITABLE

Logs can be rewritten

Records that live on your own servers can be changed — so they prove nothing to an auditor, an LP, or a court.

03 · LIABILITY

"Trust us" is the exposure

With no independent proof, every automated decision is a liability waiting for a subpoena you can't answer.

How it works

Seal. Bind. Anchor.

IRL runs as a gateway between your agent and the exchange. Three steps turn an automated decision into permanent, checkable evidence.

01 — SEAL

Before the order

Your agent submits its reasoning — model, intent, size, risk checks. IRL runs pre-execution policy, then seals the snapshot with SHA-256. Nothing trades until it passes.

02 — BIND

After the fill

The exchange returns a transaction ID. IRL computes final_proof from the sealed reasoning and the tx — binding intent to what actually executed.

03 — ANCHOR

Every day

All seals roll into a daily Merkle root, anchored to Bitcoin via OpenTimestamps. The record becomes tamper-evident against everyone — including us.

Trust without trust

You don't have to trust us.

Every audit vendor claims a tamper-proof log. Only IRL hands the proof to the other side — the auditor, the regulator, the LP — and lets them check it with zero access to our servers.

Open-source verifier. Frozen spec. Reimplement it in any language.
Check any proof bundle offline, against Bitcoin.

We can't fake it. We can't rewrite it. Neither can you. That's the entire design.

Why us

We run our own signal on this discipline.

IRL is built by the team behind MacroPulse.

MacroPulse is a macro-regime signal we run in production — Ed25519-signed and publicly verifiable, every day. Proving our own market calls is how we learned how hard real proof is. IRL's Layer 2 can bind each decision to that signed market truth. See the public track record →

Built for what's coming

Auditability is becoming law.

The direction of travel is one-way: automated decisions must be explainable and provable. IRL maps to the specific requirements each framework asks for — not vague "auditability."

MiFID II · RTS 6
Algorithmic trading records & pre-trade controls
EU AI Act
High-risk system logging & traceability
SEC 15c3-5
Pre-trade risk controls on market access
DORA
Operational resilience & evidence trails
Pricing

Start proving in a day.

Self-host in your own VPC. Per-agent pricing, unlimited calls. Verification is free for everyone, forever — your auditors never need an account.

L2
IRL Audit Platform
Prove what the market was. Signed truth + anti-replay.
$1,200 / agent / mo
Min 3 agents · 99.9% SLA
  • Everything in L1, plus:
  • Signed Ed25519 heartbeats — anti-replay
  • MacroPulse MTA integration (or your own)
  • Forensic replay of any historical trade
  • Read-only compliance dashboard
Get L2 →
L3 · Coming soon
Sovereign Gateway
Prove without revealing. For alpha you can't expose.
Enterprise
TEE + ZK · quoted per engagement
  • Everything in L2, plus:
  • TEE execution — hardware-attested enclave
  • ZK compliance proofs — prove without disclosure
  • Dedicated support · quarterly review
Talk to us →

Evaluating for a fleet? Book a technical demo → · Fleet discounts from 6+ agents.

For developers

Try it now. No signup.

Three demo agents are pre-seeded. Run a full authorize → bind flow in the interactive API, then verify the proof bundle yourself.

# POST /irl/authorize — seal the reasoning before the order
{
  "agent_id": "00000000-0000-4000-a000-000000000001",
  "model_hash": "sha256-hex-of-model",
  "action": { "Long": 1.5 },
  "asset": "BTC/USD",
  "notional": 64875.00
}
# → { "trace_id": "…", "reasoning_hash": "sha256…", "authorized": true }

Every agent decision, provable.

Start with the sandbox, self-host in a day, and hand your auditors evidence they can verify without trusting you.