Now available globally

Give AI agents cards — without giving up control

Virtual cards with real-time policy enforcement. Every transaction is approved, blocked, or escalated for human review — before money moves.

Publisher policy active User controls this agent
AI Agent
OpenAI API
$20
within daily budget
Virtual Card
single-use · merchant-locked
Policy
authorization-time check
User
delegated approval
APPROVED
Every authorization decision is written to an immutable log.
Live scenario

See an AI purchase get approved in real time

An agent initiates the payment. Shatale enforces policy, requests approval, and executes — all in seconds.

1 Agent 2 Shatale 3 User Done
AI Agent · ShopBot
ShopBot
Searching
Amazon
Nike Air Max 90
$89.00
Shatale · Intercepted
Merchant allowlist
Amount under limit
! Approval required
User Console · Approval required
Agent
ShopBot
Merchant
Amazon
$89.00
Payment authorizedPolicy-enforced · User-approved

“Agents can pay — with control built in.”

The gap

AI agents can initiate purchases. But users still need control — and publishers need a safe way to enable it.

That leaves engineering stuck between two bad options: give agents broad access to payment credentials, or keep humans in every transaction. Neither scales.

01

Shared cards create uncontrolled risk

Team cards and stored credentials don't give agents bounded permissions or clear ownership. One rogue workflow touches the entire budget.

02

After-the-fact controls are too late

If policy runs after authorization, the money already moved. Agent payments need controls at the exact moment of approval — not as a reconciliation step.

03

No audit trail means no approval to ship

Finance and compliance need to see what the agent attempted, what policy was checked, and who approved exceptions. Without this, autonomous payments can't go to production.

What Shatale provides

A payment control plane for agent-initiated spend

Shatale combines virtual card issuing, authorization-time policy enforcement, and human approval workflows — so AI agents can transact without bypassing finance controls.

A card per agent or workflow

Issue isolated virtual cards instead of reusing shared payment credentials. Each card is scoped to a specific agent, purpose, or customer account — nothing bleeds across.

Policy checks before approval

Allow or block transactions by merchant, MCC, amount, geography, time window, and custom logic. Rules run at the authorization moment — zero latency, zero exceptions.

Escalate edge cases to a human

Out-of-policy requests can be routed to an approver with full context — merchant, amount, agent, and reason — instead of silently failing open or blocking everything.

Complete authorization audit trail

Every payment attempt, policy decision, reviewer action, and final outcome is recorded immutably. Export to your data stack via webhooks or pull via API.

Designed for every layer

Safe payments for users. Zero liability for publishers. No overhead for agents.

End user

Your agent, your limits

You decide what the agent is allowed to buy — by category, merchant, or spend amount. Anything outside those bounds is blocked before the charge happens, not discovered in a bank statement.

  • Out-of-scope purchases never reach the network
  • Escalations come to you, not to finance
  • Full spend history, readable in plain language
Publisher

Ship agents, not liability

Deploy AI that pays on behalf of customers without taking on compliance risk yourself. Policy violations are caught at the authorization layer — not discovered in reconciliation at month end.

  • Audit trail ships built into the product
  • Disputes backed by machine-generated evidence
  • Finance team gets reports, not incidents
AI agent

Transparent. No added cost.

Shatale sits at the authorization layer. One API call per payment intent — same latency, no credential rotation, no per-authorization fee. The agent gets approved, escalated, or declined. Nothing else changes.

  • No extra round trips or SDK to manage
  • Works with any payment intent or workflow
  • Approved transactions proceed without delay
Lufthansa Airlines
TravelBot · travel_standard policy
Live authorization decision
24ms
€ 890.50 EUR
Category allowed (Airlines — MCC 3000)
Amount within daily limit (€ 890 of € 2,000)
Below approval threshold (€ 500 auto-approve) → escalated to user
✓ APPROVED
auth_9xK2mP · travel_standard
Why not just give the agent a card?

Everyone gives the agent a card. We give the user control.

Capability Shatale Othersgeneral payment infrastructure Generic card issuing
Agent-scoped virtual cards
Per-agent, single-use, merchant-locked
Virtual cards; not agent-scoped
General cards, custom controls
Agent API and MCP
Native publisher API; MCP-ready
General APIs; no agent MCP layer
Not available
Delegated authorization flows
Delegate, approve, deny, confirm
Issuing controls; approval UX is custom
Not available
Agent policy engine
Merchants, MCCs, timing, risk rules
Radar + Issuing; not agent-specific
Custom webhook required
User-owned limit management
Users set daily, tx, MCC limits
Platform-set controls, not user-managed
Not native
Subscription management
Renewals, trials, pause per card
Billing for merchants, not user control
No policy context
Credential vault
Temp relay credentials; user revokes
No credential vault for logins
Not available
Publisher isolation
Hard boundaries: agents, cards, policies
Connect accounts; agent isolation custom
Custom implementation
Audit trail
Agent, policy, card, tx history
Logs available; agent audit needs build
No policy context
Emergency freeze
User freezes agent or publisher instantly
Card controls; agent freeze needs build
Manual process
Agent catalog
Verified library drives user discovery
No marketplace; publishers drive traffic
Not available
Monetization cockpit
Built-in MOR, billing, fraud, catalog
Connect + Billing; AI MOR not turnkey
Not available

If you only need card issuance, general-purpose issuing may be enough. If you need AI agents to spend with bounded permissions and approval workflows, that's what Shatale is built for.

Who it's for

Built for teams putting AI agents into real purchasing workflows

Procurement

Procurement agents

Let agents buy approved software or supplies within merchant and budget limits — without exposing a shared corporate card or requiring human approval on every purchase.

Operations

Operations agents

Give internal automation bounded purchasing power for cloud spend, API subscriptions, and tooling. Each agent operates within its own policy envelope.

Customer-facing

Customer workflows

Enable agent actions that require online payments on behalf of customers — with full auditability and customer-level approval controls baked into the flow.

Built-in controls

Skills: ready-made guardrails for agent card spending

Plug in a Skill in two clicks. No payment logic to write. Your agent gets instant, tested controls that run directly in the authorization path.

Instant block

Spend Limit

Block payments above any configured limit — per transaction, daily, or monthly. Runs before the card network approves.

limit: $500/day · window: daily
Instant block

MCC Block

Allow or block entire merchant categories — gambling, airlines, subscriptions, or any 4-digit MCC list.

mode: block · MCCs: 7995 (gambling)
Instant check

Balance Check

Prevent spending when available balance drops below your threshold. Configurable floor per agent.

deny_if_below: $10.00
Real-time webhook

Transaction Notify

Receive a webhook on every settled transaction — merchant name, amount, MCC, and card last4.

webhook_url: https://your-api.com/events
Try it — pick a scenario
Online Store
$850.00
MCC 5411 · Grocery
Skills engine
BLOCKEDExceeds $500/day limit
Developer-first

Integrate card issuing and payment controls by API

Create cards, attach policies, receive authorization events, handle approvals, and export decision logs through the Shatale API and Publisher SDK.

Virtual card issuance API

Provision and manage agent cards programmatically. Scoped to agent, workflow, or customer account.

Policy configuration

Define spending rules in JSON — MCC allowlists, merchant patterns, amount limits, geo and time constraints.

Real-time auth webhooks

Receive authorization events with full context before the transaction completes. Your policy logic, our enforcement.

Approval workflow endpoints

Accept or deny escalated transactions programmatically or via the end-user console. Configurable timeouts.

Decision log export

Pull the full authorization lifecycle via REST or stream events to your data warehouse.

Travel StandardACTIVE
Categories allowedAirlines · Hotels · Car rental
Daily limit€ 2,000
Approval required above€ 500
On violationEscalate → human
Attached to: TravelBot · SaaS Bot
Try it now

Connect Shatale to your IDE

No signup needed. Install our MCP server, connect to Claude Desktop or Cursor, and ask your AI to create agents with spending controls.

Works with: Claude DesktopCursorWindsurfClaude Code
Terminal
$ npx shatale-mcp-server
✓ Shatale MCP Server started (guest mode)
Then ask your AI:
→ "Create a shopping agent with €1000 budget"
→ "Simulate buying shoes for €150 at Nike"
→ "Block gambling and alcohol categories"
Live simulation

Watch an authorization happen

Pick a scenario. The engine decides in real time — and when approval is needed, you're in control.

Procurement Bot · running workflow
Merchant
Office Supplies Co.
$650.00
Budget used $0 / $10,000
Auth latency:
Shatale console
Waiting for transaction…
Simulated · No real transactions · All data is fictional
Common questions

Everything you need to know

When you give an agent a card, the agent becomes the cardholder — it can spend anything, anywhere, with no user awareness. Shatale flips this: the user delegates to the agent within defined limits. The user gets real-time approve/deny notifications, pre-sets MCC restrictions and amount caps, and sees the full audit trail. The agent spends. The user stays in control. This is the difference between "hope the agent behaves" and "delegate with confidence."

No — that's the point. Shatale is a publisher API. You integrate in one sprint: API call to issue a delegated virtual card, policy attached, user consent captured. We handle the card issuing, compliance, KYB/KYC, fraud monitoring, and chargeback handling. Your team builds the UX. We handle the payments infrastructure.

Three layers of protection. First, the policy engine blocks anything outside the MCC whitelist or amount cap — before the transaction even processes. Second, transactions above a configurable threshold require real-time user approval (with a countdown timer). Third, full audit trail means you always know exactly which policy decision was made, why, and when. And any card or agent can be frozen instantly from the publisher dashboard or user console.

Yes — Shatale is global. We issue US commercial virtual cards (Visa/Mastercard) with global merchant acceptance. Publishers in any geography can integrate via API. EU deployment is also supported. We are not EU-only.

We are currently in early access — free for all publishers. Get access now, integrate, and start testing with real users at no cost. Pricing will be introduced later as we scale. Early partners lock in the most favorable terms.

It depends on you — a basic working integration can be done in minutes. We provide API docs, an SDK, and a publisher sandbox you can explore right now. Most publishers have a working prototype in hours and a production-ready integration in one sprint (1–2 weeks). The compliance pack (KYB docs, liability framework, integration checklist) is provided upfront so your legal team can review in parallel — no waiting on us.

Safer than alternatives. B2B2C users come through your platform — they're already verified, contextualized users, not cold-start consumers. The policy engine enforces MCC whitelists, amount caps, velocity limits, and time windows at authorization time — before money moves. We provide chargeback handling and fraud monitoring. Publishers set the risk parameters; Shatale enforces them on every transaction.

All of them. Shatale is model-agnostic and framework-agnostic. We have an MCP server (Model Context Protocol) for direct integration with Claude, GPT-4o, and other MCP-compatible agents. For other frameworks, the publisher API works with any agent that can make an HTTP call.

Still have questions? Talk to us →

Early access

Talk to the team building global payment infrastructure for AI agents

We're working with a small group of design partners. If you're building AI agents that need online purchasing capability, we'd like to hear from you.

No spam. We'll respond within 48 hours.

Application received

We'll be in touch within 48 hours.