Agents can shop.
Trust is the hard part.
Every major payment network is racing to support AI agents. But authorisation was designed for humans. Paygent is the trust layer between your agent and the checkout — verifiable intent, spending guardrails, prompt-injection defence, and audit trail included.
Why now
The agent economy is live. The trust infrastructure isn't.
AI-driven retail spend projected for 2026 — nearly 4× year-on-year
MCP, ACP, Visa TAP, X402 launched in under 12 months. Zero interoperability.
No jurisdiction has enacted agent-specific payment liability regulation yet
Trust infrastructure
Six layers that make agent payments safe
Not a payment processor. The verification, authorisation, and audit layer that sits between your agent and any checkout.
Verifiable Intent
Prove every agent was authorised
Each purchase carries a signed authorisation chain — linking the transaction to the original user instruction, agent reasoning context, and defined scope. Fraud, impersonation, and liability disputes resolved instantly.
Spending Guardrails
Limits at every dimension
Define per-transaction, daily, category, vendor, or monthly caps. Guardrails enforce silently below threshold and route to a human approver above it — with zero custom glue code.
Prompt Injection Shield
Blocks manipulation before it pays
Detects and intercepts prompt injection attempts — where a malicious vendor or page tries to override agent instructions and redirect funds — before any transaction is executed.
Protocol Agnostic
MCP, ACP, TAP, X402 — all covered
The agentic commerce standard war is real: MCP, Stripe ACP, Visa TAP, and Coinbase X402 are all live. Paygent translates across all of them so your stack never bets on the wrong horse.
Human-in-the-Loop
Approval routing that stays out of the way
High-value or out-of-policy purchases pause and push to a designated approver — via Slack, email, or in-app — then resume execution immediately on confirmation.
Immutable Audit Trail
Every action, forever accountable
Full audit log of agent identity, intent hash, authorisation scope, approval path, and transaction outcome. Dispute resolution in seconds, not days.
How it works
From instruction to trusted transaction
Four steps with trust enforced at every handoff — not bolted on after.
Agent queries providers
Consumer agent connects to the Paygent routing layer. Provider catalogs, pricing, SLAs, and policies are returned in a normalised format — regardless of underlying protocol.
Intent is signed & verified
Paygent generates a verifiable intent record: original instruction hash, agent identity, authorisation scope. Prompt-injection scanner clears the request before funds are touched.
Guardrails checked in real time
Spending guardrails run instantly. Below threshold: transaction executes in milliseconds. Above threshold: paused and routed to the human approver you configured.
Payment executes + receipt logged
Payment completes via connected gateway. Receipt, intent record, agent ID, and approval chain are written to the immutable audit log and returned to both parties.
Buyer-side agents
Real purchasing decisions,
safely automated
Agent books same-day plumber, electrician, or cleaning service from approved local vendors within configured budget.
Purchase market data pulls, analyst reports, or document processing from specialist firms on demand.
Agent manages subscription renewals, seat upgrades, and add-on purchases within pre-approved tool budgets.
Autonomous purchasing agent reorders inventory, books freight, or escalates to human when vendor is unknown.
Agent books flights, hotels, and ground transport within company travel policy — approval triggered above threshold.
Provider side
Be discoverable by agents
Register once. Expose your catalog, pricing, SLAs, and payment intents through Paygent's normalised interface — instantly reachable by agents using any protocol.
FAQ
Common questions
What problem does Paygent solve?
Traditional payment authorisation assumes a human is present — with a card, PIN, or biometric. When an AI agent initiates a purchase, none of those signals exist. Paygent provides the missing trust layer: cryptographic proof that the agent was authorised to act, within defined spending limits, for a specific purpose.
Which agent protocols does Paygent support?
Paygent is protocol-agnostic. It supports MCP (Anthropic), ACP (Stripe/OpenAI), Visa TAP, and Coinbase X402 — so your agent stack works regardless of which standard your providers or customers use.
How does the spending guardrail system work?
You define limits at any granularity: per-transaction, per-day, per-category, per-vendor, or cumulative monthly. When a purchase would breach a limit, Paygent pauses execution and routes the request to a human approver before any funds move.
What is verifiable intent and why does it matter?
Verifiable intent is a signed, auditable record that links every purchase to the original user instruction, the agent's reasoning chain, and the authorisation scope — proving the agent acted within its mandate. This protects you from prompt-injection attacks, agent impersonation, and liability disputes.
Who is Paygent for?
Paygent serves two sides of the agent economy: businesses deploying consumer agents that need to purchase services within defined budgets, and service providers who want to be discoverable and safely transactable by AI agents without building custom integrations per protocol.
Build with Paygent
Your agents are ready to buy.
Is your stack ready to trust them?
Join the builders and service providers shaping the infrastructure of the agent economy. Early access is open now.