The credential says the agent is authorised.
AIIAN proves the action is admissible.
AIIAN is the Action Prover for autonomous AI. Every proposed action is evaluated against identity, risk budget, regulatory limits, and execution environment — before any state changes. Default-deny. Cryptographically signed. Independent of any agent framework.
Patent Pending · UK Intellectual Property Office · Applying to the FCA Digital Sandbox
01
Agents are already writing code, moving money, and making commitments on behalf of organisations. The question is no longer whether the AI can do this. It is whether the action has reached a valid, attestable state — and whether there is a cryptographic proof that the evaluation happened before the state changed.
Regulatory context
ESMA's February 2026 supervisory briefing identified a gap between agent authorisation and pre-execution action validation in automated financial systems. AIIAN is designed for this gap. Built for FCA-regulated environments.
Activity ID
a3f7c2d1e9b4…Operational Status
● LiveRecent Activity
Action
FX instruction · EUR/GBP · €4,200,00002
Every consequential action is evaluated against a defined set of conditions — identity, risk limits, coverage, regulatory constraints, and execution environment. If any condition fails, execution does not proceed. Not deferred. Not flagged for later. Not released.
Verify before mutate. No valid proof, no release.
Reference implementation
Production governance shouldn't be the first time your agent meets AIIAN. The Sandbox is a full reference implementation of the convergence protocol — dispute resolution, arbitration, re-evaluation, and release — running against your agent before it touches real systems.
Works with any provider. The same protocol that governs production is the protocol you test against. There is no difference.
Arbitration produces a candidate.
Only AIIAN Core can validate and release.
Requested by
Authorised reviewer
Records exported
2,847
04
Runtime toolkits secure behaviour within a single agent's environment. AIIAN operates above them — across the full execution path. It determines whether all required controls, limits, and attestations are in place before release. Authorisation. Risk budget. Remediation path. Signed evidence. No release until the state is valid.
Framework-neutral
AIIAN sits above agent frameworks, not inside them. The same pre-execution governance layer — regardless of which model your agents run on.
Drop a single hook into your Claude Code setup. Every tool call — Bash, Write, git push — is evaluated against your policy before it runs. Zero prompt changes required.
Wrap any agent tool — OpenAI function call, Google ADK tool use, or Claude agent SDK — with the AIIAN evaluation layer. One policy profile governs all of them.
Every decision — allow, deny, escalate — is written to a tamper-evident Ed25519-signed canonical event. Ready for regulatory review, underwriter verification, or internal audit.
Agents ask AIIAN before they act. Actions are denied by default — only explicitly authorised operations within defined risk budgets are released. Policy is never inside the prompt.
DORA, EU AI Act, MiFID II, SR 11-7 — compliance flags are attached to every decision event automatically based on the target environment. Built for finance, healthcare, and energy.
Multiple agents share a governed risk budget across frameworks and counterparties. When one agent consumes capacity, others see the updated ceiling in real time. Shared governance across the entire network.
03
Each release — and each denial — is written to a tamper-evident Ed25519-signed audit record at the moment of decision. Not reconstructed after the fact. Cryptographically linked, immutable, and ready for regulators, underwriters, or counterparties.
Operating Limit
£ 850
Status
Within defined parameters
Readiness
Confirmed
05
Multiple agents share a governed risk budget across systems and counterparties. When one agent consumes capacity, all others see the updated ceiling immediately. No agent can exceed its authorised limit — not by misconfiguration, not by prompt injection, not by concurrent execution.
Capital deployed
£ 124k
Parameters
Applied
Accredited participants only. Invitation required.
Intent
git push origin main --force06
Connect AIIAN via a PreToolUse hook or SDK wrapper. Every consequential operation — file writes, deploys, API mutations, financial instructions — is evaluated against policy before it executes. Allow, deny, or hold for attestation. Policy is never inside the prompt.
Claude Code · Claude Agent SDK · LangChain · AutoGen · CrewAI · Microsoft Agent Framework · Azure AI Foundry · any MCP-compatible stack.
For regulated financial institutions
AIIAN ensures that autonomous transactions converge to a valid, attestable execution state before they enter external systems — across counterparties, settlement rails, and regulatory constraints.
No valid state, no release.
Execution request
Convergence check — 7 conditions
Release token · Ed25519
tok_a3f7…valid 5 minCompliance Templates
Pick a certified template for your industry. Policies, spend limits, audit hooks, and compliance flags — all pre-configured. Instantiate in seconds.
All templates include a signed compliance certificate, tamper-evident audit trail, and Default-Deny policy hooks.
Custom templates available on Enterprise.
Get started
AIIAN is in active pilot with regulated financial institutions. If your agents execute high-consequence transactions — FX instructions, credit approvals, settlement operations, document commitments — we want to talk.
Pilot access is by invitation. We respond to every request personally. No sales funnel.
Initial focus: European financial institutions operating under DORA, Basel III, and MiFID II.
Patent Pending · UK Intellectual Property Office