Authority API for AI agents

Give an agent enough authority to be useful.

Stateful, versioned spending authority: what an agent may spend, with whom, and when — evaluated against the budget it has already consumed, before any payment moves. Deterministic, never an LLM, and enforced on whatever rail executes.

mnd8t controls what AI agents are allowed to do — and creates a verifiable record of what they actually did, and whether the mandate was fulfilled.

Pay a supplierSettle an invoiceTop up a metered API
Decision latency
0.31s
LLMs in the decision path
0
Customer keys held
0
Receipts per decision
1signed
The missing layer

Identity proves who an agent is. Payment infrastructure lets it transact. Neither decides what it is authorised to do.

Card networks, commerce protocols, agent identity standards and stablecoin rails are being built quickly and well. They answer who is calling, what instrument it holds, and how value moves. None of them answers whether this particular payment is within the authority a person actually granted — because that depends on the budget already spent, this counterparty's history, and whether a human would have wanted to be asked.

That is the layer mnd8t is built for. We also maintain a public reference on the whole stack — the Agentic Authority Stack, a dated tracker, and analyses of AP2, ACP, Visa TAP and Mastercard Agent Pay.

Read the intelligence hub
GovernanceWho answers when it goes wrong?
CommerceHow do agent and merchant talk?
RailHow does value actually move?
CredentialWhat instrument does it present?
IdentityWho is this agent, and who is it acting for?
AuthorityWhat is this agent allowed to do?mnd8t is built for this layer

Four steps between an agent and your money.

01

Write the mandate

Approved counterparties, per-transaction limits, daily and monthly budgets, permitted purposes, operating hours. Versioned and immutable once published.

02

Run it in shadow

Replay real transactions and see what mnd8t would have decided. Zero risk to the workflow you already have. Switch to enforcement when the policy has earned it.

03

The agent asks before it pays

A deterministic engine answers APPROVE, REJECT, or ESCALATE in seconds, with the full reason set and a rule-by-rule trace.

04

You execute, mnd8t proves

An approved action receives a one-time signed authorisation artifact your own system verifies, claims, and executes with credentials mnd8t never touches.

Authorise · Execute · Verify

An approved payment is not a fulfilled mandate.

mnd8t doesn't stop at permission. It creates a traceable link between what an agent was authorised to do, what it actually did, and the evidence that the intended outcome occurred.

An agent authorised to buy 100 units for no more than £10,000 can pay £9,700, on time, to the right supplier, entirely within its mandate — and take delivery of 80. Every authority system stops one step before that fact. A mandate that records only what was permitted cannot tell you whether delegating it worked.

authorise
Mandate
What you delegated, and what you delegated it for. Versioned and immutable.
authorise
Policy decision
Deterministic, stateful, before any money moves.
authorise
Authorisation
A one-time signed artifact your executor claims.
execute
Execution
Your rails, your keys. mnd8t never holds either.
verify
Evidence
What settled, what was invoiced, what actually arrived.
verify
Outcome
Fulfilled, or not — against the objective as agreed.
See approved ≠ fulfilled in the sandboxEvidence is customer-supplied · no carrier, merchant or ERP integrations
Evidence receipt

Every decision is a document you can verify without us.

Who delegated the authority, which policy version applied, why the decision was made, who approved any exception, and what actually executed — Ed25519-signed and hash-chained. Change one timestamp and verification fails.

Three possible outcomes
✓ APPROVE⚠ ESCALATE✕ REJECT
int_7bd910 · rule trace
mandate/published1.00
schedule/within-window1.00
budget/daily-limit0.88
!counterparty/first-seen0.79
approval/autonomous-limit0.62
verdict · escalate to a human£1,200.00

Start in shadow. Enforce when it's earned.

A sandbox key, an agent, and a mandate. Most teams reach their first decision in under 30 minutes.

Get a sandbox key
No custody · No keys · No order routing