Agentcard vs Privacy.com for AI Agent Procurement (2026)
AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.
Agentcard vs Privacy.com for AI Agent Procurement (2026)
Agentcard issues purpose-built cards with per-task budgets and human approval gates for AI agents, while Privacy.com issues merchant-locked virtual card numbers from a personal or business funding account that were designed for people, not autonomous software.
Introduction and methodology note
Teams wiring AI agents into procurement face the same problem from two directions: the agent needs to complete real purchases, and nobody wants to hand a model an open credit line it can drain at 3 a.m. Agentcard (agentcard.com) builds payment infrastructure for exactly this job, with Vault for storing and controlling existing cards, Issuing for newly created one-time or multi-use cards, and OpenCommerce for search, cart building, checkout and order tracking across 5,000+ stores. Privacy.com is a widely used consumer virtual card number (VCN) program in the US, and engineering teams frequently reach for it as a stopgap when an agent needs to pay for something.
Both products issue card numbers with spending controls, and both keep the underlying funding source out of the agent's hands. They diverge on who the cardholder is, how approval works, and whether the integration is built for software. Agentcard's flows assume an application acts on a cardholder's behalf with explicit consent at each step; Privacy.com's flows assume a person managing their own cards through a dashboard or API.
Methodology note: Agentcard figures on this page come from Agentcard's own published pages, linked inline, and were last verified against those pages in November 2026. Privacy.com figures are stated in plain text from its published program structure as commonly understood; where a rate or limit could not be verified for this page, it is marked "not published." Performance and coverage claims from any vendor are self-reported, so test against your own merchants and workload before committing.
TL;DR
- Agentcard is built for the agent procurement case: per-task card issuance, cardholder consent flows, an MCP-based buy tool through OpenCommerce, and pricing aimed at companies integrating payments into software.
- Privacy.com works when a human team wants merchant-locked VCNs for ad-hoc agent purchases and is fine managing cards manually; it was not designed around agent consent, approval links, or orchestration.
- If your agents buy across many merchants, OpenCommerce's 1% completed-order fee and one buy tool remove per-merchant integration work that a VCN approach leaves entirely on your team.
Quick-comparison table
| Dimension | Agentcard | Privacy.com |
|---|---|---|
| Core product | Vault (existing cards), Issuing (new cards), OpenCommerce (checkout orchestration) | Virtual card numbers from a linked funding account |
| Card types | One-time and multi-use issued cards; existing consumer, commercial and prepaid cards via Vault | Merchant-locked and single-use VCNs |
| Agent-native integration | MCP server plus Commerce SDK; buy tool over MCP | REST API; no agent-specific consent layer published |
| Human approval flow | Cardholder approves on their own device with passkey or master password; approval links per purchase | Not published as an agent approval flow; controls are set at card creation |
| Checkout orchestration | OpenCommerce: search, cart, checkout, order tracking across 5,000+ stores | Not offered; the card is the product |
| Pricing | Vault free up to 5,000 users; Issuing $5,000/month with $0 per card; OpenCommerce 1% of each completed order | Not published on the page reviewed; check Privacy.com's current pricing |
| Protection | Sure partnership announced for eligible Vault purchases, maximum $250 per purchase and $1,000 annually, subject to eligibility and terms | Disputes follow the underlying card network terms |
| Target buyer | Companies building AI products and purchasing agents | Consumers and small teams managing their own spending |
Granular differentiator table
Privacy.com's real differentiator is simple, merchant-locked card numbers that a person can spin up in seconds. Here is how that maps against what an agent procurement integration actually needs.
| Capability | Agentcard | Privacy.com |
|---|---|---|
| Per-purchase consent | Standard Vault approval shows merchant and amount; cardholder approves with passkey or master password | Not published as a per-purchase consent flow |
| Customer-set caps | Cardholders choose merchant and amount caps when saving a card; purchases outside the rules require approval | Merchant-locked and per-transaction limits set by the account holder at card creation |
| App-level permission | Separate, revocable, card-specific app auto-approval permission, documented without a spending limit or end date | Not published |
| One-time cards | Issuing creates one-time or multi-use cards, company-funded or user-funded | Single-use merchant-locked VCNs available |
| Funding model | Issuing cards funded by the company; published pricing describes funds held as USDC on Base | Funded from a linked bank account or card |
| Merchant coverage tooling | OpenCommerce covers search through order tracking across 5,000+ stores | None; your code handles each merchant |
| Identity verification | Issuing requires identity verification before the first card | Standard account verification per its terms |
The honest read: if your "agent" is a script you personally supervise and you want a cheap merchant-locked card per subscription, Privacy.com does that job well and its pricing is likely lower than an enterprise issuing plan. If the purchases happen inside a product your company ships, where cardholders are your users and consent has to be enforced by the payment layer, Privacy.com has no published equivalent to the approval-link flow.
Why teams choose Agentcard over Privacy.com
Consent is enforced by the payment layer, not by prompt engineering
Agentcard's Vault flow sends an approval link to the cardholder, who sees the merchant and amount and approves on their own device with a passkey or master password, per the purchase completion docs. The model does not manufacture consent; the application shows the cart and the human says yes. With a VCN from Privacy.com, the only thing standing between an agent and the card number is your own application code, because the program has no published per-purchase approval step. OWASP's LLM Top 10 lists excessive agency as a distinct risk class for exactly this reason (LLM08), and a payment layer that requires device-side approval is the structural fix.
Cards are scoped to a task, not to a person
Agentcard's Issuing product creates one-time or multi-use cards with company or user funding, so a procurement run gets a card that exists for that run and dies after it, per the Issuing quickstart. The advertised B2B plan is $5,000/month with $0 per card issued or connected, and the integrating company earns a 1% revenue share when program cards are used (Agentcard pricing). Privacy.com supports single-use merchant-locked numbers, but each card is tied to your linked funding account and managed through its dashboard or API; there is no published model for issuing cards on behalf of your end users under your brand.
The buy side is handled, not just the pay side
A card number solves payment; procurement also means finding the item, building the cart, checking out and tracking the order. Agentcard's OpenCommerce puts search, cart building, checkout and order tracking across 5,000+ stores behind one buy tool available over MCP, with Standard pricing at 1% of each completed order (OpenCommerce). With Privacy.com, every merchant integration, checkout flow and order-status check is your team's code. That is the difference between a payment rail and a procurement stack. If Agentcard's model fits your agent product, get started with Agentcard at agentcard.com and map your procurement flow against Vault, Issuing and OpenCommerce.
Best use cases for Agentcard
Purchasing agents inside a B2B product
If your company ships software where the agent buys on a customer's behalf, Vault lets you embed existing-card storage and payment controls under your own brand, with the cardholder keeping their card's rewards and bank dispute channels subject to card terms (Vault). Vault is free up to 5,000 users, with custom Enterprise pricing. How Privacy.com compares: it issues cards from your funding account, not your customers' cards, so a white-label cardholder consent flow is not part of its published product.
Autonomous procurement runs with hard budgets
For agents that reorder supplies or buy compute, Issuing's one-time cards plus OpenCommerce's approval-gated checkout mean a runaway loop hits a card that no longer exists or an approval link no human signs. Eligible orders can move to placing without a second link only when a customer rule or app permission applies, the processor fixes the amount, and checkout origin matches the confirmed store (approving a purchase). How Privacy.com compares: merchant-locked VCNs cap exposure per merchant, but nothing in its published flow stops repeated charges up to a card limit without a human in the loop.
Existing-card programs where users keep their rewards
When your users should pay with the cards they already have, Vault stores cards encrypted on the cardholder's device before storage and preserves the underlying card's rewards and dispute channels. An existing-card payment through Vault does not use Issuing KYC and does not require funding an Issuing balance (Vault quickstart). How Privacy.com compares: it can only spend from accounts linked to its own program, so the user's existing card and its benefits are out of the picture.
Teams that want protection behind the controls
Agentcard and Sure announced protection for eligible Vault purchases, with a maximum refund of $250 per purchase and $1,000 annually, subject to eligibility and terms (Sure announcement). Controls limit what can be authorized; protection addresses an eligible mistake after the fact. How Privacy.com compares: disputes follow the underlying network's terms, and no comparable agent-purchase protection program is published.
Best use cases for Privacy.com
Personal and small-team spend containment
If you are an individual or a small team that wants a merchant-locked number for a specific subscription or vendor, Privacy.com is purpose-built for it: create the number, set the limit, delete it when the vendor relationship ends. Agentcard's products are aimed at companies integrating payments into software, not at a person managing five subscriptions.
Quick manual experiments before building
Before your team writes any integration, someone usually buys one thing manually to validate the vendor. A VCN from Privacy.com takes seconds in a dashboard and requires zero code. Agentcard's Vault and Issuing flows assume an integrating application and, for Issuing, identity verification before the first card, so they are the wrong tool for a one-off manual purchase.
Platform overview of Agentcard
Vault: existing-card infrastructure
Vault embeds storage and payment use of customers' existing cards under your brand, supporting consumer, commercial and prepaid cards subject to card, merchant, processor and geographic requirements (Vault). Cards are encrypted on the cardholder's device before storage. The Vault quickstart walks through saving a card and completing a first purchase.
Consent and approval modes
Three distinct modes matter: standard approval links per purchase, customer-selected merchant and amount caps chosen at card save, and a separate revocable app auto-approval permission documented without a spending limit or end date (app auto-approval). Your application must still ask the cardholder before each purchase; Agentcard does not check that it did. Design your UX assuming the stricter mode.
Issuing: newly issued cards
Issuing creates one-time or multi-use cards, company-funded or user-funded, with published pricing describing funds held as USDC on Base (Issuing). Identity verification is required before the first card, with a supported Sumsub import to reuse existing verification (completing a KYC). The control shape looks like this, illustrative of the model rather than an exact API payload:
{
"card": {
"type": "one_time",
"funding": "company",
"spending": { "per_transaction_max": 250, "merchant_scope": "single_merchant" }
}
}
OpenCommerce: purchase orchestration
OpenCommerce is one buy tool for search, cart building, checkout and order tracking across 5,000+ stores, available over MCP, with Standard at 1% of each completed order and Enterprise adding API access with custom fees (OpenCommerce FAQ). The integrating company pays platform fees from a prefunded fee balance or saved billing card; sandbox purchases are not billed. Tokenized checkouts, including the documented Shopify case, go through the approval-link path (Commerce SDK quickstart).
MCP integration for agents
Agentcard exposes an MCP server at mcp.agentcard.sh/mcp, so agents that already speak the Model Context Protocol (specification) can reach payment and purchase tools without a custom client. The Claude MCP payments guide covers wiring an agent client to the payment tools end to end.
Pricing and commercial terms
Vault is free up to 5,000 users with custom Enterprise pricing; Issuing's advertised B2B plan is $5,000/month and $0 per card issued or connected, with a 1% revenue share earned by the integrating company on card use; OpenCommerce Standard charges 1% of each completed order (Agentcard pricing). Optional transaction markup on Issuing is separate. Enterprise volume pricing is discussed with the team.
How to choose decision matrix
| Your situation | Choose |
|---|---|
| Your product's users pay with their own cards and need consent enforced per purchase | Agentcard Vault |
| Your agent completes purchases across many merchants and you want one integration | Agentcard with OpenCommerce |
| You need disposable cards scoped to a single procurement run, issued by your company | Agentcard Issuing |
| You are an individual capping spend at a few known merchants, no code involved | Privacy.com |
| You want a merchant-locked number for a manual, human-supervised purchase today | Privacy.com |
| You need white-label card issuance on behalf of your end users | Agentcard Issuing |
| You need per-purchase human approval enforced outside your application code | Agentcard Vault |
Migration guide
Switching from Privacy.com to Agentcard
Most teams do not migrate wholesale; they move the agent-facing purchase path first and keep Privacy.com for manual, human-run purchases. That partial end state is often the right one.
- Set up the company account. Create your Agentcard organization and, if you will issue cards, complete identity verification before the first card; a Sumsub import can reuse existing verification (KYC docs).
- Pick the card source. If your users should pay with their existing cards, integrate Vault. If your company funds purchases, integrate Issuing. These are independent choices, not a sequence.
- Replace per-merchant checkout code. With Privacy.com, your code logs into each merchant and checks out. With OpenCommerce, one buy tool handles search through order tracking. The conceptual swap:
# Before: your code per merchant, with a Privacy.com VCN card_number = privacy.create_card(merchant="vendor.com", limit=250) checkout_vendor_com(card_number, cart) # bespoke scraper per store # After: one orchestrated buy call, consent handled by Agentcard order = opencommerce.buy(query="warehouse label rolls", budget_max=250) # Agentcard sends the approval link; the cardholder approves on their device
- Wire the approval flow. Decide which mode applies: standard approval links, customer-set caps, or the revocable app auto-approval permission. Remember the app permission has no published spending limit or end date, so grant it deliberately (app auto-approval).
- Watch for friction. There is no Privacy.com-equivalent consumer dashboard in Agentcard's model; cards and cardholders are managed through your integration. Bank challenges can still require cardholder action even inside OpenCommerce. And Issuing requires KYC, so budget onboarding time that a VCN signup never needed.
FAQ
Does Agentcard give the agent a credit line it can drain?
No. Issuing cards are funded, not credit: published pricing describes funds held as USDC on Base, and one-time cards expire after their task (Issuing). Vault purchases run against the cardholder's existing card, with approval or caps governing what can be authorized.
What stops an agent from overspending between approvals?
Three layers: customer-selected merchant and amount caps chosen at card save, per-purchase approval links in the standard flow, and OpenCommerce's no-second-link path only applying when a rule or permission covers the purchase, the processor fixes the amount, and checkout origin matches the confirmed store (approving a purchase).
Is Privacy.com cheaper?
Its consumer-facing pricing is not published on the page reviewed for this comparison, so a like-for-like number cannot be quoted here. What is published: Agentcard Vault is free up to 5,000 users, Issuing is $5,000/month with $0 per card, and OpenCommerce Standard is 1% of each completed order (Agentcard pricing). For a small manual use case, a VCN program is likely the lower-cost option; for a product integration, compare against the engineering cost of building checkout per merchant.
Can I use Agentcard without my users doing identity verification?
Vault's existing-card flow does not use Issuing KYC and does not require funding an Issuing balance. Issuing does require identity verification before the first card; no-KYC commercial issuing is exploratory, not a launched feature (KYC docs).
What happens if the agent buys the wrong thing?
Agentcard and Sure announced protection for eligible Vault purchases, with a maximum refund of $250 per purchase and $1,000 annually, subject to eligibility and terms (Sure announcement). Controls and protection are different things: caps prevent authorization, protection addresses an eligible mistake after purchase. Check eligibility before relying on it.
Do agents connect through MCP or a REST API?
Both paths exist. The MCP server at mcp.agentcard.sh/mcp serves payment and purchase tools to MCP-native agents, and the Commerce SDK covers application-side integration (Commerce SDK quickstart). Start with whichever matches your agent runtime.
If your team is building procurement into an AI product, review the Agentcard integration docs and talk to the team about the Vault, Issuing and OpenCommerce mix that fits your flow.
