# When AI Agents Pay and Sellers Monetize: Building x402 Apps on AWS — Anil Nadiminti, AWS

Anil Nadiminti, AWS

[Talk page](https://ai.engineer/talks/qTZirYu9pr0-when-ai-agents-pay-sellers-monetize-building) · [Video](https://www.youtube.com/watch?v=qTZirYu9pr0)

## When AI Agents Pay and Sellers Monetize: Building x402 Apps on AWS

Anil Nadiminti explains how x402 turns a payment into access to a resource, how AgentCore Payments separates spending from the agent loop, and how AWS WAF lets sellers price AI requests at the edge.

### A human paywall stops an autonomous task

[0:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=12s)

A news portal works until the reader hits a paywall. A human can enter credit card details, choose a weekly, monthly or annual subscription, and obtain access. An autonomous agent encounters the same obstacle while completing a task, but the next step becomes a request for help: someone must supply payment details or an API key. Anil Nadiminti, a Senior Solutions Architect at AWS, starts with this familiar example because it exposes where autonomy breaks. The agent can find the content; it cannot finish obtaining it.

The urgency comes from a change in who requests resources. Nadiminti puts bot traffic above human traffic and says AI agents account for 95 percent of that bot traffic. The recording does not establish the population or measurement method behind those figures, so they serve as his framing rather than a universal traffic estimate. His forecast—about a billion agents performing tasks by 2027, with 60 percent of enterprises using agentic workflows—extends the same premise: more requests will arrive from software trying to complete work independently.

The news publisher has an uncomfortable choice:

- **Block bots:** Blocking can forfeit AI discovery, citations and potential licensing partnerships. Content that an agent cannot reach is harder for it to incorporate into an answer.
- **Allow unpaid access:** Large numbers of requests create infrastructure costs, while unrestricted reuse can weaken attribution and control over the content.

Paid machine access offers another option: let the agent obtain the resource and let the publisher earn from the request.

### Buyers need spending controls; sellers need payment at the edge

[4:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=252s)

Agent ecommerce joins two different requirements. On the buy side, an agent needs access to premium or licensed resources, a wallet and the ability to make small payments. An enterprise also needs to prevent that agent from going on a spending spree. On the sell side, a publisher needs to understand the requesting bot and charge for its activity without rebuilding the origin—the infrastructure that stores and serves the content. A shared machine-to-machine payment protocol lets these sides meet at the request.

The proposed change is from approving each purchase to setting conditions under which purchases may happen. Humans move from being in the transaction loop to supervising it, or staying outside it once limits are configured. Pay-per-use and pay-per-execution can then follow the agent's work at request speed, rather than requiring a new human subscription workflow whenever it encounters a paid resource.

Small payments make the fee structure decisive. Nadiminti uses a card-payment example with a 25-cent minimum fee plus 2.5 percent. For a purchase worth a tenth of a cent, the fixed fee alone is 250 times the purchase: $0.25 ÷ $0.001 = 250. The percentage fee barely matters beside that floor. Charging separately for each tiny resource request therefore requires a payment path whose overhead fits the size of the purchase.

### x402 makes payment part of the resource request

[6:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=401s)

**x402** builds on HTTP status code `402`, reserved for “Payment Required.” Nadiminti presents the protocol introduced by Coinbase as a way for machines to negotiate access through HTTP. Return to the news portal: instead of sending the agent into a human checkout, the server responds that payment is required. The client chooses a payment method and sends payment authorization back to the server.

Authorization is followed by verification and settlement. The server uses a facilitator to verify the payment authorization and complete the transaction. In the flow described here, the server releases the content after settlement completes on chain. The observable change is simple: the resource request that previously stalled now reaches the content response through a payment exchange.

What must happen between the first request and delivery of the content? The diagram separates the HTTP exchange from the facilitator's work. Payment authorization starts the transaction process; completed settlement is what precedes content delivery.

This is the force of the phrase “payment is the … credential”: the purchase itself enables access, without setting up a subscription or a separate API key for that resource. Nadiminti describes no protocol fee and nominal merchant gas fees. His description of “zero wait time” means an automated request-speed experience, rather than literal zero latency; the later figures include a 200-millisecond average settlement time on Base.

The protocol is also presented as extensible and open to implementation. Nadiminti dates its introduction to May 2025 and describes Linux Foundation open governance, with support from Coinbase, AWS, Google, Stripe, Anthropic, Cloudflare and Circle. That shared protocol supplies the connection between the AWS buyer and seller services introduced next.

From payment-required response to content
The server uses a facilitator for verification and settlement, then returns the requested content.
- Client — Requests a resource, chooses a payment method and sends authorization.
- Resource server — Returns Payment Required and receives payment authorization.
- Facilitator — Verifies authorization and completes the transaction.
- On-chain settlement — Completes before the server delivers content in this flow.
- Client → Resource server (1. Request resource)
- Resource server → Client (2. Payment Required)
- Client → Resource server (3. Payment authorization)
- Resource server → Facilitator (4. Verify and complete payment)
- Facilitator → On-chain settlement (5. Settle transaction)
- On-chain settlement → Resource server (6. Settlement completed)
- Resource server → Client (7. Return content)

### AgentCore Payments gives the buyer a wallet and a budget

[9:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=551s)

**AgentCore Payments**, part of Bedrock AgentCore, supplies the buyer's payment machinery. The service described in the recording supports Coinbase wallets and a Stripe preview wallet option, with payment connectors orchestrating transactions. It supports x402 at the time of the presentation; support for additional protocols is planned. Protocol independence is a design goal, while x402 is the concrete path shown.

A payment session constrains spending programmatically. It can specify a maximum amount and an expiry expressed in minutes. Nadiminti's example is permission to spend $5 over 30 or 60 days. The amount limits how much the session can spend; the expiry limits how long that permission lasts. Built-in observability is intended to let developers trace the activity across the stack.

The agent continues using tools, MCP servers and other resources to complete the user's task. When one of those requests returns `402`, AgentCore Payments handles the transaction and tells the agent that settlement has happened. The agent can then continue toward its response. Applied to the opening example, the news paywall becomes a payment event handled by the configured service, rather than a reason to ask the user to enter card details mid-task.

Discovery is another part of the buyer's path. AgentCore Gateway exposes internal APIs through MCP, making them available as agent tools. Its integration also gives AgentCore Payments access to a Coinbase discovery service that Nadiminti describes as offering more than 10,000 endpoints available for transactions. Finding a purchasable resource and paying for it are thus separate capabilities connected through the gateway and payment service.

### Keep keys and payment execution outside the agent loop

[11:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=701s)

The wallet integration does not give the agent its private keys. Imported secret keys are stored in a secure token wallet protected by KMS, and the agent cannot access them. That distinction matters: an agent may request a paid resource without receiving the secrets that enable the payment infrastructure to transact.

Payment execution is also decoupled from the agent's reasoning loop. Skills and inputs can be poisoned by malicious actors, so the payment path runs in a deterministic layer responsible for connectors, orchestration, limits and third-party wallet integration. The model can continue choosing tools and pursuing the task, while spending controls operate outside its nondeterministic decisions. This design contains payment authority; it does not make poisoned agent inputs harmless.

Where does a resource request cross into payment execution? The diagram shows the `402` response triggering the separate payment layer, with session controls and protected wallet keys attached to that layer. The return path communicates settlement to the agent rather than passing private keys into its loop.

The separation also supports reuse of existing agent models and frameworks. Nadiminti describes payment handling as a layer that can operate without changing the agent's code. In the narrated demo, an agent discovers a secured resource, AgentCore Payments uses an already integrated wallet, and the transaction completes. The change from the opening paywall is the removal of the manual payment interruption, with the wallet and controls configured beforehand.

Agent reasoning and payment execution
Resource access triggers payment handling, but wallet secrets remain in the protected payment infrastructure.
- Agent loop — Uses tools and resources with the chosen model and framework.
- Paid resource — A tool or resource request can return HTTP 402.
- AgentCore Payments — Handles orchestration and wallet integration in a separate deterministic layer.
- Payment session — Sets a maximum spend and expiry.
- KMS-protected token wallet — Stores imported wallet secrets; the agent cannot read the private keys.
- Agent loop → Paid resource (Request through tools)
- Paid resource → AgentCore Payments (402 triggers payment handling)
- Payment session → AgentCore Payments (Constrain session spending)
- KMS-protected token wallet → AgentCore Payments (Support imported wallet integration)
- AgentCore Payments → Agent loop (Notify settlement; task continues)

### WAF prices the request before it reaches the origin

[14:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=880s)

The seller needs enough information about a request to choose a price. Nadiminti describes AWS Web Application Firewall bot detection as recognizing more than 650 bot types, verifying bots by signature and identifying intent, such as model training or retrieval-augmented search. Identity and intent answer different questions: which organization is requesting the content, and what kind of use is the request associated with? The talk does not explain the intent-classification method or its accuracy, so intent-based pricing depends on that classification working.

**WAF AI traffic monetization** applies that context at the edge. With CloudFront, the content distribution network, a publisher can attach WAF and configure monetization through the console or infrastructure as code. Nadiminti also describes applying WAF to internal APIs exposed through AgentCore Gateway as MCP tools. In both cases, the seller adds payment handling in front of an existing resource rather than moving the resource into a new application.

The seller-side flow detects and categorizes the bot, identifies its intent, verifies it and uses x402 to monetize the request. The service is presented as requiring no SDK or origin changes, with publishers retaining 100 percent of revenue and no transaction or subscription fees. Those claims concern the monetization service; they do not remove the infrastructure costs that motivated the seller's dilemma or the gas fees described in the protocol discussion.

Pricing can use several independent dimensions:

- **Path:** `/blog`, `/research` and an API endpoint can have different rates because they expose different resources.
- **Identity:** A verified bot from an organization with an existing commercial relationship can receive a different price from an unverified bot. Anthropic is the example partner in the talk.
- **Intent:** Training access can have a different rate from search access.

WAF rules combine these conditions with AND/OR logic. A seller can therefore price a request using several attributes together, rather than charging every bot the same amount.

The news portal can now distinguish human readers, verified bots and unverified bots. Human access might remain free or carry its own price, while bots follow different payment rules. Revenue dashboards aggregate results by bot and show the paths being accessed, giving the publisher a way to inspect which resources and requesters contribute to its revenue model.

### The transaction examples extend beyond content

[18:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1120s)

Paid news access is only one use. The closing examples include LLM inference, compute, web scraping, research agents, agent-to-agent transactions and monetized MCP services. Each gives an agent a resource or capability it can purchase while performing work. The same request-and-payment pattern can therefore apply to executing a tool as well as reading a page.

For the preceding 12 months of Coinbase marketplace activity, Nadiminti reports $50 million in transaction volume across 170 million transactions. He also gives an average settlement time of 200 milliseconds on Base and a cost of about a tenth of a cent per transaction. These figures describe the cited activity and settlement setting, rather than a benchmark of the AWS services. They make the economic target concrete: many small purchases need both low payment overhead and short settlement delays.

Payments sit within the larger Bedrock AgentCore ecosystem. Builders can bring their own model and framework, add memory, use managed knowledge bases and web search, expose internal APIs through MCP, and evaluate agent performance. AgentCore Runtime hosts the agentic application at scale; Nadiminti describes each request as running in its own isolated micro virtual machine. The final architecture keeps these jobs distinct: the runtime serves the agent, the gateway connects resources, and the payment layer handles purchases when access requires them.

## Key takeaways

- x402 turns a resource request into a payment exchange: the server returns Payment Required, receives authorization, uses a facilitator for verification and settlement, then delivers content.
- AgentCore Payments places session budgets, expiry, wallet integration and payment execution outside the agent's reasoning loop. Imported private keys remain in a KMS-protected store the agent cannot access.
- Seller-side WAF rules can combine resource path, verified bot identity and classified intent to set prices at the edge without changing the origin.
- The economic target is payment overhead small enough for individual resource purchases. A 25-cent fixed fee is 250 times a tenth-of-a-cent purchase; the closing examples extend these purchases to inference, compute and MCP tools.

## Related talks

- [Why Your AI Agent Needs a Wallet: USDC and Nanopayments — Harshal Bhangale, Circle](https://ai.engineer/talks/xKzU_3riL6s-why-your-ai-agent-needs-wallet-usdc) — A companion topic for the wallet and small-payment economics behind paid resource requests.
- [x402 isn’t good (yet) — Jan Curn, Apify](https://ai.engineer/talks/h6mi88VrPtQ-x402-isnt-good-yet-jan-curn-apify) — A contrasting perspective on the same protocol, useful alongside this presentation's adoption case.
- [Building Deterministic Infrastructure for Non-Deterministic AI Agents](https://ai.engineer/talks/APh1Vx0oLmQ-building-deterministic-infrastructure-non) — Extends the architectural question raised by keeping payment execution outside a nondeterministic agent loop.

## Slides

### When AI Agents Pay and Sellers Monetize

Anil Nadiminti introduces building AWS apps to support agent e-commerce.

[0:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=20s) [Title slide: Accelerating Agentic Commerce with AWS.](https://assets.ai.engineer/sites/aiecode2025/sha256/07644c7a09d9f751211d4b76faabcff07baa15eda6620de7ca10a1059e924dcd.webp)

When AI Agents Pay and Sellers Monetize:

Accelerating Agentic Commerce
with AWS

### You Have an Amazing Product for Humans

Nadiminti uses The Economist’s homepage to introduce a familiar news portal, where readers access news and may encounter a paywall.

[0:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=32s) [The Economist homepage illustrates a news product designed for human readers.](https://assets.ai.engineer/sites/aiecode2025/sha256/8511740de3eff8c19a0194a10b202a113deb2b0282514addb79cd6177f2edd0a.webp)

You have an amazing product for humans

The world in brief | 17 min ago
Catch up: Tensions rise in Hormuz; Iran’s foreign minister to visit China

Insider →

Interface text:
The Economist
Subscribe
The Economist Pro
Log in
Menu
Weekly edition
World in brief
United States
China
Business
Finance & economics
Europe
Asia
Middle East
Americas
Artificial in

### Business Model

The Economist’s paywall offers digital subscriptions on annual and four-week plans. The speaker uses it to illustrate how readers pay for access to news content.

[0:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=46s) [The Economist subscription page shows digital plans and their benefits.](https://assets.ai.engineer/sites/aiecode2025/sha256/e3779cebcea7e2237cccb6b88354873a4ab69bfeaf487033e557a1c5ebd4fcd3.webp)

Business model

Get unlimited access to The Economist

Now with Insider—weekly shows giving you unprecedented access to our senior editors

Digital
Digital + Print

BEST VALUE
Annual — $3.88/week
Save 46% vs 4 week plan
Billed as $289 $202.30 for first year, then auto-renews at $289 annually

4 week — $7.25/week
Includes 1-week free trial
Auto-renews at $29 every 4 weeks when free trial ends

Cancel auto-renewal at any time
Continue to checkout

Award-winning journalism with informed analysis
Fact checked, data driven, in-depth coverage
Full access to Insider—video debates with our senior editors

Get a global perspective
Stay informed on the world’s most important developments with our incisive coverage

Interface text:
The Economist
The Economist Pro
Log in
United States

### AI Agents Are Consuming Content

The slide contrasts subscription paywalls with bot traffic: it reports that bots account for 51% of all web traffic, while AI scraping bots account for 95% of web traffic at some businesses.

[1:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=67s) [The slide reports that bots account for 51% of web traffic and AI scraping bots account for 95% of traffic at some businesses.](https://assets.ai.engineer/sites/aiecode2025/sha256/ec60640b7625c8ce90abb0428e4021f4fde90ac263ac8961d5f12608407a1caa.webp)

AI agents are consuming content – but the economics don’t work

51%
of all web traffic is now automated bots — surpassing humans for the first time

Thales 2025 Bad Bot Report

95%
of web traffic at some businesses now comes from AI scraping bots

Bloomberg Law, Aug 2025

### The Rise of Autonomous Agents

The timeline moves from assistants answering questions to autonomous agents that self-direct. It projects 1B+ agent tasks per day and 60% of enterprise workflows agentic by 2027.

[1:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=93s) [A timeline progresses from assistants and copilots to agents and autonomous agents, with forecasts for agent tasks and enterprise workflows by 2027.](https://assets.ai.engineer/sites/aiecode2025/sha256/0f1ee91e706c74685627c698e92518ec65de1e42209f8d95bbda3ff73fce4634.webp)

The Rise of Autonomous Agents
From chatbots that answer questions to agents that take action —> autonomously.

2023
Assistants
Answer questions, summarise text

2024
Copilots
Draft code, suggest next actions

2025
Agents
Plan, use tools, execute multi-step tasks

Now →
Autonomous Agents
Operate independently & self-direct

1B+ Agent tasks run per day by 2027
60% of enterprise workflows agentic by 2027

### Agents Can't Pay

Agents stall at paid resources until a person supplies approval, card details or credentials, interrupting autonomous workflows.

[2:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=140s) [Two panels describe agents stalling at paid resources and human involvement creating payment friction.](https://assets.ai.engineer/sites/aiecode2025/sha256/51cc4a149eff6182f7403352e03af115654fc04273f7b648ebc1d4b3fd3c466a.webp)

Agents Can't Pay — And That's Holding Everything Back

Agents Stall

AI agents hit paywalled APIs, premium services, and rate-limited resources — and stop. They can't proceed without human approval.

Manual Friction

Every payment requires a human in the loop: approvals, credential sharing, card details. Kills autonomous workflows entirely.

### Today you have options

Content sellers face a tradeoff: blocking bots can cost discovery and licensing opportunities, while allowing them can increase infrastructure costs without programmatic compensation.

[2:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=168s) [Two panels compare the stated drawbacks of blocking all bots and allowing all AI bots.](https://assets.ai.engineer/sites/aiecode2025/sha256/ea1b34f57c7ce1bddc0c654f355bbf815bd0fcc2d8e2816db227bef66b966be5.webp)

Today you have options.

BLOCK ALL BOTS
Lose AI-powered discovery
Miss partnership and licensing opportunities
Become invisible in AI search results

VS

ALLOW ALL AI BOTS
Infrastructure costs surge with no return
Content is used by AI interfaces without programmatic compensation mechanism
Subscription and ad revenue erode

### The Rise of Autonomous Agents

The slide traces a progression from assistants answering questions to autonomous agents, then proposes agents that pay, get paid and settle value autonomously.

[3:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=236s) [Timeline progresses from assistants, copilots and agents to autonomous agents and a proposed next stage of agents that transact.](https://assets.ai.engineer/sites/aiecode2025/sha256/80629ee6b6e009c02cb540e56985770b8a8521118599efb3dfc75e9648ecaa12.webp)

The Rise of Autonomous Agents
From chatbots that answer questions to agents that can transact autonomously

2023 — Assistants
Answer questions, summarise text

2024 — Copilots
Draft code, suggest next actions

2025 — Agents
Plan, use tools, execute multi-step tasks

Now → Autonomous Agents
Operate independently & self-direct

Next → Agents that Transact
Pay, get paid & settle value autonomously

### Agentic commerce

Agentic commerce lets autonomous AI agents discover resources, make payments and access content on behalf of users or organizations.

[4:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=245s) [Defines agentic commerce as autonomous AI agents executing purchases and payments on behalf of users or organizations.](https://assets.ai.engineer/sites/aiecode2025/sha256/5e8a8d3ced20eb82ea3441b031226f56cd251696e15292e2458cdc5c23d6bd0a.webp)

Agentic commerce describes an emerging form of e-commerce in which autonomous artificial intelligence (AI) agents independently execute purchasing and payment processes on behalf of users or organizations.

### The Two sides of Agentic Commerce

Agents need payment tools and spending guardrails; content sellers need to identify and charge bots at the edge. Nadiminti frames a standard machine-to-machine payment protocol as a shared need.

[4:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=260s) [Side-by-side panels list payment and guardrail needs for agents and monetization and enforcement needs for content sellers.](https://assets.ai.engineer/sites/aiecode2025/sha256/696b56df5be50d2ce84746b20b9c036150de5eac06764e07ec1c53a744458b3e.webp)

The Two sides of Agentic Commerce
The agentic economy needs both buyers AND sellers to work simultaneously

BUY SIDE
Agents that need to pay

AI agents need premium APIs, paywalled content, licensed data to complete tasks
No way for agents to hold wallets, authorize payments, or settle autonomously
Per-request micro-payments ($0.001–$10) don't fit monthly SaaS models
Enterprises need guardrails: budgets, approvals, audit trails for agent spend

SELL SIDE
Services that deserve to get paid

Content owners see billions of bot requests scraping their data — unpaid
Only options today: block bots entirely, or let them scrape for free
No way to identify agent intent, verify willingness to pay, or price per request
Need edge enforcement — monetize at scale without touching application code

Both sides need the same thing: a standard protocol for machine-to-machine payments at the edge

### The Two sides of Agentic Commerce

Buyers want agents to pay for content and APIs without a human approving every transaction; sellers want to earn from AI traffic without rebuilding their stack or negotiating with every AI company.

[5:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=337s) [Buyer and seller statements describe paying for content and APIs and earning from AI traffic without extensive manual intervention.](https://assets.ai.engineer/sites/aiecode2025/sha256/8f0d4035869673386c074595149ea9a8f026f89ec58cd04575a4d4a03918b235.webp)

The Two sides of Agentic Commerce

BUYER SIDE
“My agent needs to pay for content and APIs — without a human approving every transaction”

SELLER SIDE
“I want to earn from AI traffic — without rebuilding my stack or negotiating with every AI company”

### Business models are being rewritten

Nadiminti envisions agents replacing subscription and one-size-fits-all pricing with real-time micropayments per call or execution, while humans move from approving each transaction to overseeing agents or stepping out of the loop.

[5:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=351s) [Today’s subscription models and human approvals are contrasted with agent micropayments and pay-per-use pricing.](https://assets.ai.engineer/sites/aiecode2025/sha256/4705a87580e44212cc7a948fc6228ee36cfb9a50c09bb0259c4938ed86ff5101.webp)

Business models are being rewritten

Today
• Subscription models
• Human-in-the-loop
• One-size fits all pricing

The Future
• Billions of agents making micro-transactions
• Fractions of a cent per-call billed in real time
• Pay-per-use, per execution

aws
© 2026, Amazon Web Services, Inc. or its affiliates. All rights reserved.

### Today’s payment rails were built for humans

A $0.001 API call illustrates the problem: the $0.25 minimum card fee alone is 250 times the call’s stated cost, before the 2.5% charge.

[6:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=380s) [An example compares a $0.001 API call with a card transaction fee of at least $0.25.](https://assets.ai.engineer/sites/aiecode2025/sha256/fe760cfac454e9386027eb1fb5e9c96846f305648cdd9dfe401b16fab0eeb88b.webp)

Today's payment rails were built for humans

The math doesn't work for agents— not agents making millions of tiny calls

25¢ + 2.5%
The minimum fee on every card transaction

What an agent actually needs to buy

A single API call
costs the provider $0.001

But the card fee is
minimum per transaction $0.25+

That's 250× the price of the thing being bought

### HTTP status codes

The slide moves from familiar 200, 404 and 301 responses to 402 Payment Required, a code it says was reserved in 1997 and is now being put to use.

[6:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=416s) [Cards describe HTTP 200 OK, 404 Not Found, and 301 Moved Permanently.](https://assets.ai.engineer/sites/aiecode2025/sha256/42789a81c647632029568c55a1485b8558114cddfd93dd9afa902902629c3a24.webp)

A brief history ....
Every HTTP request gets an answer

HTTP has a status code for every outcome

200
OK
Request succeeded — here's your content.

404
Not Found
The resource doesn't exist here.

301
Moved Permanently
Redirecting you to the new location.

[7:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=423s) [The slide highlights HTTP 402 Payment Required alongside dimmed 200, 404, and 301 cards.](https://assets.ai.engineer/sites/aiecode2025/sha256/5a2b3f9a6e3f53dac5a30a82b39cf0d0aa82e5ed32a236b6815f47e3bbc4b29d.webp)

A brief history ....
Every HTTP request gets an answer

HTTP has a status code for every outcome

200
OK
Request succeeded — here's your content.

404
Not Found
The resource doesn't exist here.

301
Moved Permanently
Redirecting you to the new location.

402
Payment Required
Reserved in 1997 — never implemented. Until now.

### Introducing x402

A server returns 402 with payment requirements. The agent pays on-chain, and the server verifies payment and fulfills the request, automating machine-to-machine micropayments.

[7:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=429s) [An overview describes payment requirements, an agent payment, server verification, and automated fulfillment.](https://assets.ai.engineer/sites/aiecode2025/sha256/4e7b80486a52045635e176b628f4906a815dda0a68f2ba433d9f0845c96f0557.webp)

Introducing x402

x402 is an open protocol that extends HTTP to support machine-to-machine micropayments.

Server responds with 402 + payment requirements
Client (the agent) sees what's needed & cost
Agent pays via crypto / stablecoin on-chain
Server verifies payment, fulfills the request
All automated — zero human intervention

### x402 Payment Flow

After a 402 response, the client sends payment authorization. The server uses a facilitator to verify and settle it on-chain, then returns the response.

[7:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=466s) [Sequence diagram showing a client request, payment authorization, verification, blockchain settlement, and server response.](https://assets.ai.engineer/sites/aiecode2025/sha256/0f326dd7f4113c74ec5958f8f27ba4f3c9d971e7965bf123cb7828fe549164d8.webp)

x402 Payment flow

CLIENT   SERVER   FACILITATOR   BLOCKCHAIN

1 GET /api
2 402 - Payment required
3 Select payment method & payload
4 Include Header: X-PAYMENT: b64 payload
5 /verify
6 Verification
7 do work to fulfill request
8 /settle
9 Submit tx w/ sig to usdc contract
10 Tx confirmed
11 Settled
12 return response w/ X-PAYMENT RESPONSE

### What makes x402 compelling for developers and businesses

x402 has no protocol fees, though payment network fees remain. The speaker says payment serves as the credential to access content; anyone can build on or extend x402.

[8:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=500s) [Slide presenting zero protocol fees, zero wait, and zero friction as x402 benefits.](https://assets.ai.engineer/sites/aiecode2025/sha256/ebbb77fb8e0d6ebbd1b0edb5cb598c2128dd54dcc7a4c45bde5d1f641d2ae05b.webp)

What makes x402 compelling for developers and businesses

Zero protocol fees
x402 is free for the customer and the merchant—just pay nominal payment network fees

Zero wait
Money moves at the speed of the internet

Zero friction
No accounts or personal information needed

Zero centralization
Anyone on the internet can build on or extend x402

[8:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=507s) [Four visible x402 benefits: zero protocol fees, wait, friction and centralization.](https://assets.ai.engineer/sites/aiecode2025/sha256/7587084a79dcc6fe85939fefc3cdd0dcaca60144f4e6b4ba5b7fb9a0f296a612.webp)

What makes x402 compelling for developers and businesses

Zero protocol fees
x402 is free for the customer and the merchant—just pay nominal payment network fees

Zero wait
Money moves at the speed of the internet

Zero friction
No accounts or personal information needed

Zero centralization
Anyone on the internet can build on or extend x402

Zero restrictions
x402 is a neutral standard, not tied to any specific network

### x402 Key Milestones

The timeline traces x402 from Coinbase’s May 2025 introduction through its April 2026 donation to the Linux Foundation and the May 2026 launch of Amazon Bedrock AgentCore Payments.

[8:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=519s) [Timeline showing x402's introduction, its contribution to the Linux Foundation and the AgentCore Payments launch.](https://assets.ai.engineer/sites/aiecode2025/sha256/d0cb3da354c425c6e2d33cf9b59d7033fcb085e1576aef044339195d6ea9238e.webp)

x402 Key Milestones

ANNOUNCEMENT
Linux Foundation is Launching the x402 Foundation and Welcoming the Contribution of the x402 Protocol
THE LINUX FOUNDATION

Timeline:
May 2025 — Coinbase introduces x402 protocol
Apr 2026 — Donated to Linux Foundation
May 2026 — Amazon Bedrock AgentCore Payments Launch

Backed by Coinbase, AWS, Google, Visa, Stripe, Cloudflare, Circle, Anthropic

18

### What we heard from developers

Developers want buyer-side agents to have wallets, real-time settlement and pay-per-use endpoints, with budgets and guardrails. They also want observability across the stack to trace what happens under the hood.

[9:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=550s) [Five developer needs: wallet support, payment orchestration and settlements, pay-per-use endpoints, budgets and guardrails, and observability.](https://assets.ai.engineer/sites/aiecode2025/sha256/5bad2c384cae3f2d0a79e20a67c209366b7442fde28a0b44b3bb81f157e5df71.webp)

What we heard from developers...

WALLET SUPPORT
PAYMENT ORCHESTRATION & REAL-TIME SETTLEMENTS
READY TO USE PAY-PER USE ENDPOINTS
BUDGETS AND GUARDRAILS
OBSERVABILITY

aws
© 2026, Amazon Web Services, Inc. or its affiliates. All rights reserved.

### Amazon Bedrock AgentCore Payments

AgentCore Payments is available in preview, enabling AI agents to autonomously discover, authorize, and execute payments using wallets from Coinbase and Stripe.

[9:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=570s) [AgentCore Payments preview announcement, in partnership with Coinbase and Stripe.](https://assets.ai.engineer/sites/aiecode2025/sha256/b2318b670482832796ca4064a21843d62415501108f227228cc2315d97c790a9.webp)

Amazon Bedrock
AgentCore Payments

NEW

Enable AI agents to autonomously discover, authorize, and execute payments with a few lines of code

AVAILABLE IN PREVIEW

in partnership with
coinbase  stripe

### Payment Limits

A payment session associates available spend with a maximum spend amount and an expiry in minutes. The maximum can be set programmatically.

[10:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=642s) [A payment session connects to available spend, maximum spend, and expiry time limits.](https://assets.ai.engineer/sites/aiecode2025/sha256/e04e48e55e912ab01d58c7d808e931e37d83b34f497da04b5ee2962a18b3be5e.webp)

Payment Limits

Payment Session
Available Spend Amount
Max Spend Amount
Expiry Time In Minutes

aws
© 2026, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark.
23

## Full transcript

[0:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1s) [music]

[0:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=12s) Hello all, welcome to uh the agent

[0:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=14s) e-commerce track and uh I'm Anil Lminti.

[0:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=17s) I'm a senior solutions architect here at

[0:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=19s) AWS. Uh I'm I'm here to talk to you

[0:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=21s) today about how AWS is innovating and uh

[0:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=25s) how you can build apps on uh AWS to

[0:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=29s) support the agent e-commerce. So uh

[0:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=32s) welcome to the session. Uh just to get

[0:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=34s) you started, let me set the stage with

[0:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=36s) something that you're very familiar

[0:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=37s) with. Uh imagine that you are your

[0:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=40s) organization is building uh a news

[0:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=43s) portal like this, right? So you're all

[0:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=45s) familiar with something where you're

[0:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=46s) accessing the news content and then

[0:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=48s) suddenly you hit a payw wall, right? So

[0:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=50s) this is where uh you pull out your

[0:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=52s) wallet or you try to uh figure out how

[0:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=55s) to make the payments, set up your uh

[0:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=57s) credentials, access keys in the sense

[0:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=59s) that you you make a credit card

[1:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=61s) transaction weekly, monthly or annual

[1:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=63s) subscription and then get started to

[1:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=65s) access the content. Right? So this is

[1:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=66s) all the content that is behind a payw

[1:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=68s) wall. But uh what we see now is that uh

[1:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=71s) much of the traffic that is actually

[1:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=73s) being uh sent to these uh uh portals now

[1:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=76s) on the internet is all coming from bots.

[1:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=78s) We see that we at a infliction point

[1:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=80s) where uh the bot traffic is more than

[1:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=83s) the human traffic right. So it's just uh

[1:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=85s) actually in fact surpassed that and uh

[1:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=88s) 95% of that bot traffic is coming from

[1:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=90s) AI agents. So uh essentially we are also

[1:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=95s) looking at the rise of autonomous agents

[1:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=97s) right so where we all started using LLMs

[1:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=101s) asking questions asking for

[1:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=102s) summarization being able to uh get help

[1:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=105s) with using them as co-pilots getting

[1:47](https://www.youtube.com/watch?v=qTZirYu9pr0&t=107s) them to do agentic work to uh do

[1:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=109s) multi-step tasks and now we're in the

[1:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=111s) phase of autonomous agents where agents

[1:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=114s) are using the reasoning powers of large

[1:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=116s) language models to complete a task and

[1:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=118s) completing a task means that uh it it

[2:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=120s) has to go do whatever you're asking it

[2:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=123s) to do. And that's kind of where we are

[2:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=124s) in the uh journey. And we see that by

[2:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=127s) 2027 about a billion agents will be

[2:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=130s) running uh performing tasks and 60% of

[2:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=133s) the enterprises will already be uh using

[2:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=135s) agentic workflows. So uh what happens

[2:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=138s) when agents hit these pay walls that uh

[2:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=140s) we just saw? When agents hit the pay

[2:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=142s) walls, they stall, they can't operate

[2:24](https://www.youtube.com/watch?v=qTZirYu9pr0&t=144s) and you see those messages that hey I

[2:26](https://www.youtube.com/watch?v=qTZirYu9pr0&t=146s) cannot access content. Uh right? So at

[2:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=148s) that point humans get in the loop. They

[2:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=151s) try to enter and put the credit card

[2:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=154s) details or API keys do the transactions

[2:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=157s) for the AI agents. But all of that is

[2:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=159s) manual friction, right? So essentially

[2:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=161s) bringing in a human in the loop. So

[2:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=163s) autonom autonomous agents actually break

[2:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=165s) [clears throat] at that point where the

[2:47](https://www.youtube.com/watch?v=qTZirYu9pr0&t=167s) uh friction is now building up. So uh

[2:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=170s) now sellers of the content have uh you

[2:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=173s) know couple of options right. So block

[2:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=175s) all the bot traffic but by blocking all

[2:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=177s) the traffic they lose this AI powered

[2:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=179s) discovery they miss this uh partnership

[3:02](https://www.youtube.com/watch?v=qTZirYu9pr0&t=182s) licensing options and AI also now

[3:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=186s) supports citations right so the

[3:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=187s) responses so they lose all of that uh

[3:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=189s) powered citations as well if they can't

[3:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=190s) sell the content essentially they lose

[3:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=193s) uh revenue generating options and if you

[3:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=195s) allow the bots to access those uh uh the

[3:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=198s) content what it means is that you know

[3:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=200s) hundreds of thousands of bots or

[3:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=202s) millions of bots could be hitting your

[3:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=203s) uh uh infrastructure which means that

[3:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=205s) the infrastructure costs will also raise

[3:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=207s) and uh you need to be able to support

[3:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=209s) all of that right so you also uh when

[3:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=211s) you allow bots to access content you

[3:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=213s) lose attribution the IP itself right so

[3:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=215s) because content is now freely available

[3:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=217s) so uh both these decisions are probably

[3:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=221s) not not a good option they're not ideal

[3:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=223s) so there should be another ideal option

[3:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=225s) where uh you would want to have your AI

[3:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=228s) agents being able to get and pay for the

[3:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=230s) content that they are looking for and

[3:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=233s) monetize on

[3:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=234s) So uh now we look at the next phase of

[3:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=237s) uh uh rise in autonomous agents where

[4:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=240s) agents should be able to transact and

[4:02](https://www.youtube.com/watch?v=qTZirYu9pr0&t=242s) make uh discover other agents resources

[4:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=245s) and essentially make payments. Right? So

[4:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=247s) this is the definition of agent

[4:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=248s) e-commerce where AI agents can

[4:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=250s) essentially discover uh you know it's a

[4:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=252s) form of e-commerce where autonomous

[4:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=254s) agents can uh discover independently and

[4:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=256s) make those uh uh settlements and then

[4:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=258s) access content. So let's look at uh the

[4:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=262s) agent e-commerce the two sides of agent

[4:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=263s) e-commerce the buy side and the sell

[4:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=265s) side. So uh when we talk about the buy

[4:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=267s) side the agents are making these

[4:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=269s) transactions and on the sell side the

[4:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=271s) sellers of the content are trying to

[4:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=273s) monetize on the content. So on the buy

[4:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=275s) side when you look at things AI agents

[4:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=277s) want to access these premium paywalled

[4:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=279s) content licensed content they want to be

[4:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=281s) able to hold wallets which they do not

[4:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=283s) have the option today and they want to

[4:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=285s) be able to make these microtransactions

[4:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=286s) you just heard in the prior talk as

[4:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=288s) well. But enterprises when they come to

[4:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=290s) this point they want more guardrails and

[4:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=292s) they do not want agents to go on

[4:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=294s) spending spree. Think about it right?

[4:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=296s) Would you allow your AI agents to get

[4:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=297s) handled on your wallets or credit cards

[4:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=299s) to be able to do that transactions and

[5:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=301s) where they could go rogue as well right?

[5:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=303s) So that's what the buyer side is looking

[5:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=305s) at. And on the seller side uh the there

[5:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=308s) are again billions of transactions that

[5:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=310s) be happening with these AI bots. So

[5:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=311s) sellers really want to be able to

[5:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=313s) understand what kinds of bots are

[5:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=315s) operating uh what kinds of transactions

[5:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=317s) they're making and uh really uh do this

[5:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=320s) at the edge. The sellers don't want to

[5:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=322s) change their entire infrastructure and

[5:24](https://www.youtube.com/watch?v=qTZirYu9pr0&t=324s) origins where the content is sitting.

[5:26](https://www.youtube.com/watch?v=qTZirYu9pr0&t=326s) They want to be able to do this at the

[5:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=327s) edge without changing much of this.

[5:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=329s) Right? So there is again uh one common

[5:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=332s) thing here on the buyer side and the

[5:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=333s) seller side which is a standardized

[5:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=335s) approach or a protocol to be able to

[5:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=336s) solve for this uh machine to machine

[5:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=338s) payments at the edge. So uh bottom line

[5:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=341s) buyers are saying that they want their

[5:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=342s) agents to be able to pay for content uh

[5:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=345s) and uh not have humans approving this

[5:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=348s) and then the sellers are saying that

[5:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=349s) they want to be able to earn from the AI

[5:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=351s) traffic. So bottom line the subscription

[5:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=354s) model is going to change with humans in

[5:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=356s) the loop to becoming humans on the loop

[5:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=359s) or out of the loop and that's kind of

[6:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=361s) what we are building towards. uh the the

[6:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=363s) traditional one-sizefits model does not

[6:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=366s) work anymore because of the fact that uh

[6:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=368s) again we look at that in the next slide

[6:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=370s) where uh the transactions cost will not

[6:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=372s) really work right all of this needs to

[6:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=374s) happening at realtime speed and the

[6:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=377s) paper use and paper execution is what

[6:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=379s) the f future is going to look like so if

[6:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=382s) you're a seller you would have come

[6:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=383s) across this right so there is a 25 cent

[6:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=385s) minimum transaction fees as well as 2.5%

[6:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=389s) on top of that and all of these

[6:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=391s) microtransactions are uh you know in the

[6:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=393s) in in like a cent subsend or you know

[6:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=396s) micro cents is what we are calling them.

[6:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=398s) So if you add like a 25 cents on top of

[6:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=400s) that it's essentially like 250 times to

[6:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=402s) what you know they are essentially

[6:44](https://www.youtube.com/watch?v=qTZirYu9pr0&t=404s) paying for. So the all of this model

[6:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=406s) does not work and uh while we are trying

[6:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=409s) to solve for that a very brief history

[6:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=410s) uh of this is every HTTP call

[6:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=413s) essentially responds back uh you know

[6:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=414s) there's a response for that you've seen

[6:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=416s) 200 status codes 404 uh you know and the

[6:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=419s) 301 these are all like status codes that

[7:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=421s) you're familiar with and then there is

[7:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=423s) one status code which is 402 which has

[7:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=425s) uh not been used it was reserved for

[7:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=427s) payment required and now finally uh

[7:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=430s) Coinbase has introduced this uh as uh

[7:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=433s) transactions over 402 which is also

[7:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=435s) called as X42 where they uh you know the

[7:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=437s) protocol talks about how you can do

[7:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=439s) machine to-achine transactions uh using

[7:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=441s) this protocol right so we'll take a

[7:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=443s) closer look at that but uh what happens

[7:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=445s) within the protocol is uh you know if

[7:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=447s) you look at this uh flowchart here a

[7:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=449s) client makes a request to the server and

[7:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=451s) then the server responds back with the

[7:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=453s) payment required uh the client then

[7:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=456s) figures out what is the payment method

[7:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=457s) that it wants to operate and then it

[7:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=459s) sends the payment authorization to the

[7:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=461s) server the server then utilizes a

[7:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=463s) facilitator to complete the verification

[7:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=466s) and then also utilizes the same

[7:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=468s) facilitator to complete the transaction

[7:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=470s) and once the settlement is completed

[7:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=472s) onchain essentially the server will then

[7:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=474s) respond back with the content right so

[7:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=476s) this is what's happening under the X42

[7:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=478s) protocol I thought I'll pick one of the

[7:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=479s) protocols and just uh explain this to

[8:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=481s) you but uh why this is compelling is uh

[8:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=483s) you know essentially there is no

[8:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=484s) protocol fees uh or the fees that a

[8:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=487s) consumer is paying for uh you know these

[8:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=490s) microcent transactions and the merchant

[8:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=492s) is paying very nominal gas fees is uh

[8:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=494s) again there is zero wait time this is

[8:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=496s) happening at the speed of internet and

[8:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=499s) uh there is no friction there is no API

[8:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=501s) keys to set up no subscriptions and the

[8:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=503s) payment is the essentially the uh

[8:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=505s) credential to be able to get the content

[8:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=507s) so there is no centralization it's uh

[8:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=510s) x42 can be extended as well and you can

[8:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=512s) implement it and there are no

[8:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=513s) restrictions as well so some key

[8:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=515s) milestones here are you know it was

[8:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=517s) introduced last year uh May 2025 it

[8:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=521s) explored is not part of the Linux

[8:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=523s) Foundation under open governance and

[8:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=526s) it's backed by Coinbase, AWS,

[8:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=529s) Google, Stripe, Anthropic, Cloudflare

[8:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=532s) and Circle right so many more folks in

[8:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=535s) there that are supporting that

[8:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=536s) organizations in there. So uh from

[8:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=538s) Amazon we have also released agent core

[9:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=540s) payments uh under the bedrock uh suite

[9:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=543s) so where uh agents will make be able to

[9:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=546s) make payments and we'll go into some of

[9:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=547s) the details here. So let's talk about

[9:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=549s) the buyer side here and what is involved

[9:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=551s) right? So we uh we understood from the

[9:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=553s) developers that they really want to be

[9:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=555s) able to get this uh agents to have

[9:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=556s) wallet support. They want to be able to

[9:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=558s) have real-time settlement uh have the

[9:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=561s) budget and guardrails which enterprises

[9:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=563s) really want and observability throughout

[9:26](https://www.youtube.com/watch?v=qTZirYu9pr0&t=566s) the stack where they would want to have

[9:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=568s) uh the full stack trace of everything

[9:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=569s) that's happening uh under the hood. So

[9:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=572s) I'm excited to share with you that we've

[9:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=574s) launched agent core payments and this is

[9:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=575s) a service that allows AI agents to

[9:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=578s) autonomously discover uh authorize and

[9:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=580s) execute payments with a few lines of

[9:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=582s) code. Uh now we've launched this in

[9:44](https://www.youtube.com/watch?v=qTZirYu9pr0&t=584s) partnership with Coinbase and Stripe

[9:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=586s) where you can bring wallets from

[9:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=589s) Coinbase and Stripe preview to be able

[9:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=590s) to do these operations and we'll go into

[9:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=593s) some of the details but uh the core

[9:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=595s) capabilities to start with are wallet

[9:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=596s) support where you can bring the the

[9:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=599s) wallets from Coinbase and Stripe. you're

[10:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=601s) able to orchestrate the payments using

[10:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=603s) payment connectors and uh today we

[10:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=605s) support X42 with many more protocols to

[10:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=608s) uh you know that are in the pipeline.

[10:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=609s) The service is designed to be protocol

[10:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=611s) agnostic. So as new protocols emerge, we

[10:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=614s) are going to be adding the support for

[10:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=616s) those protocols as well. And uh you know

[10:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=618s) the the settlement is going to be

[10:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=620s) instantaneous uh instant essentially and

[10:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=623s) uh the payment limits can be set uh

[10:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=625s) which is the most important thing that

[10:26](https://www.youtube.com/watch?v=qTZirYu9pr0&t=626s) we spoke about where enterprises are

[10:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=628s) looking to put some payment limits and

[10:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=630s) guards on how these transactions can

[10:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=631s) operate. So uh observability is builtin

[10:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=635s) and uh essentially all of this uh

[10:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=637s) operates with uh you know security as

[10:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=639s) the uh layer that is operating uh the

[10:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=641s) whole model right. So with that let's

[10:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=643s) look at uh some of these uh details on

[10:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=646s) how the payments limit can be set up

[10:47](https://www.youtube.com/watch?v=qTZirYu9pr0&t=647s) right. So you can create payment

[10:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=648s) sessions where you can set the

[10:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=651s) programmatically set the maximum amount

[10:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=653s) of uh uh value that can be used for

[10:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=656s) transactions or you can also set expiry

[10:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=659s) time in minutes. think where uh you are

[11:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=661s) able to set that uh I can the agent can

[11:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=664s) actually spend maybe $5 in 30 days or 60

[11:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=668s) days right so that's kind of the

[11:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=669s) operation uh model that you can set with

[11:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=671s) many more uh you know details that are

[11:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=673s) available I'm only going over a few

[11:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=675s) features but uh let's look at what

[11:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=678s) happens on the buyer side when the user

[11:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=680s) is asking an agent to make a particular

[11:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=683s) uh you know requesting for resources

[11:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=685s) right so agent completes the request by

[11:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=688s) accessing tools MPPP servers other

[11:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=690s) resources as well. So at that point of

[11:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=692s) time if the agent uh is uh looking at

[11:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=696s) you know it it finds that the there is a

[11:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=698s) response from one of the tool calls or

[11:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=700s) requests with a 402 agent core payments

[11:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=703s) is going to handle the request uh to

[11:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=705s) complete the transaction and then let

[11:47](https://www.youtube.com/watch?v=qTZirYu9pr0&t=707s) the AI agent know that uh essentially

[11:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=709s) the settlement happened and the AI uh

[11:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=712s) agent will be able to respond back with

[11:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=714s) the users. So in this process when the

[11:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=717s) wallets are uh wallet support is

[11:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=719s) imported the the the secret keys that

[12:02](https://www.youtube.com/watch?v=qTZirYu9pr0&t=722s) you use to import the wallets actually

[12:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=724s) are stored in a secure token wallet that

[12:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=726s) is uh secured by KMS where you know

[12:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=730s) that's there. So essentially the agent

[12:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=732s) does not have access to the private

[12:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=733s) keys. This is most important uh to note.

[12:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=736s) And uh next thing is that uh agent core

[12:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=738s) payments is also integrated uh through

[12:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=740s) uh a gateway which is also part of uh

[12:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=742s) which is another service that we have to

[12:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=745s) mp5 your internal APIs. Uh through agent

[12:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=748s) core gateway uh the agent core payments

[12:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=751s) can get access to discovery service uh

[12:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=753s) in coinbase where there are 10,000 plus

[12:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=756s) endpoints that are available to transact

[12:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=758s) and then u again there is a per session

[12:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=760s) budget that we just discussed as well.

[12:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=763s) So uh there is a decoupling of agent

[12:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=766s) infrastructure and uh the payment

[12:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=768s) infrastructure by design where the agent

[12:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=770s) can operate it it in its own loop and

[12:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=772s) whenever it sees the payment the payment

[12:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=775s) uh connectors orchestration payment

[12:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=776s) limits and integration with third party

[12:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=778s) wallets can happen right so it's

[13:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=780s) important to decouple them because again

[13:02](https://www.youtube.com/watch?v=qTZirYu9pr0&t=782s) skills can be poisoned inputs for the

[13:05](https://www.youtube.com/watch?v=qTZirYu9pr0&t=785s) agents can also be uh you know poisoned

[13:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=787s) by inputs as well right so where

[13:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=789s) malicious actors could uh try to do that

[13:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=792s) so By decoupling and making this uh by

[13:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=796s) design a agents can essentially have a

[13:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=798s) secure path for these transactions and

[13:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=800s) payments do not touch the you know

[13:24](https://www.youtube.com/watch?v=qTZirYu9pr0&t=804s) undeterministic path but this is more on

[13:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=805s) a deterministic uh uh layer as well. So

[13:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=808s) why this is important is that uh again

[13:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=810s) agents uh the code of the agents does

[13:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=813s) not have to change. You can bring your

[13:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=815s) own model frameworks and then the

[13:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=816s) payment itself uh can flow through uh in

[13:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=819s) the payment uh uh layer itself. So again

[13:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=823s) the controls the policies pending

[13:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=825s) controls can be outside of the payment

[13:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=828s) stack itself and again it's this is

[13:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=830s) built to be protocol agnostic. So this

[13:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=832s) is uh one of the console screens where

[13:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=834s) it shows how you can import uh the

[13:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=836s) payment connector uh and it shows that

[13:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=839s) you know you can select the the coinbase

[14:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=841s) wallet and the stripe preview wallet

[14:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=843s) from the console and uh this is a demo

[14:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=846s) in action where we are showing how a

[14:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=848s) secure resource can be accessed. Now in

[14:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=850s) this case uh the AI agent is essentially

[14:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=854s) making a you know discovering that there

[14:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=855s) is a secure source the agent core

[14:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=858s) payments is kicking in and then it's

[14:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=860s) completing the transaction by utilizing

[14:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=863s) the wallet that is already integrated

[14:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=865s) and uh the transaction completes. Now uh

[14:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=869s) this is on the buyer side. Now let's

[14:32](https://www.youtube.com/watch?v=qTZirYu9pr0&t=872s) look at the seller side to understand

[14:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=873s) what's happening. Right? So uh again

[14:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=875s) there is a lot of bot activity that's

[14:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=877s) happening. We have released uh under uh

[14:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=880s) the uh AWS web application firewall a

[14:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=882s) feature where we have bot detection in

[14:44](https://www.youtube.com/watch?v=qTZirYu9pr0&t=884s) place. Today we detect over 650

[14:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=886s) different types of bots. Think of bots

[14:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=888s) like perplexity bot, GPD bot, cloud bot,

[14:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=891s) you know, again Google bots, right? So

[14:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=893s) there are so many bots that are out

[14:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=894s) there. So we're able to detect also

[14:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=897s) understand the intent of these bots. So

[14:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=898s) why are these bots accessing the

[15:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=900s) content? Are they accessing the content

[15:02](https://www.youtube.com/watch?v=qTZirYu9pr0&t=902s) to train their models? are they doing it

[15:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=904s) because they have to respond back to an

[15:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=906s) intent where they're uh responding for a

[15:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=908s) rag search. So we're able to identify

[15:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=910s) the intent. We're also able to verify

[15:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=912s) the bots and identify them by a

[15:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=914s) signature. So we are able to say hey

[15:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=917s) this is a verified bot. So maybe you

[15:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=919s) have uh built a relation with one of

[15:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=921s) these organizations and these

[15:24](https://www.youtube.com/watch?v=qTZirYu9pr0&t=924s) verification will allow you to have a

[15:26](https://www.youtube.com/watch?v=qTZirYu9pr0&t=926s) different pricing for the organizations

[15:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=928s) that are already verified. So we'll look

[15:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=930s) at that in a second. So there's also

[15:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=931s) real-time traffic analysis that allows

[15:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=934s) uh more uh to be customized. And I'm

[15:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=936s) also happy to share with you today that

[15:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=938s) we announced VAF AI traffic

[15:41](https://www.youtube.com/watch?v=qTZirYu9pr0&t=941s) monetization. This is a service that

[15:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=942s) allows you to monetize uh based on the

[15:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=945s) content that uh you know based on how

[15:47](https://www.youtube.com/watch?v=qTZirYu9pr0&t=947s) you can measure, verify and monetize

[15:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=949s) based on the AI traffic that is hitting

[15:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=951s) your endpoints. Now if you might be

[15:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=953s) familiar with CloudFront which is our

[15:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=955s) content distribution network you can add

[15:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=957s) a web application firewall at that point

[15:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=959s) and essentially you can moni start

[16:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=961s) monetizing uh right away and based on a

[16:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=964s) few clicks uh you can do that again

[16:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=966s) using infrastructure as code as well.

[16:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=967s) Now I also spoke about a gateway service

[16:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=970s) that allows you to expose your AI

[16:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=972s) endpoints uh that are internal use and

[16:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=975s) mcpify them. So the same web application

[16:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=977s) firewalls can be used there. So your

[16:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=979s) internal APIs can be MCPI and then you

[16:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=981s) can start monetizing as well. So what

[16:23](https://www.youtube.com/watch?v=qTZirYu9pr0&t=983s) happens during monetization? The AI

[16:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=985s) agent AI bot essentially requests for

[16:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=988s) some content. The bot context uh

[16:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=991s) understands what kinds of bots is

[16:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=993s) detecting it. It's able to detect the

[16:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=994s) bot. It's able to categorize and

[16:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=997s) understand the intent of the bot as we

[16:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=999s) discussed earlier and verify uh and

[16:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1002s) check what kind of bot is available. So

[16:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1003s) then we are able to monetize uh using

[16:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1006s) the X42 and the publishers get paid as

[16:49](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1009s) well. So important to note is that again

[16:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1011s) there is no SDK change no changes at the

[16:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1013s) origin origin. Publishers keep 100% of

[16:56](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1016s) the revenue as well and uh again there

[16:59](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1019s) is no transaction fees or subscription

[17:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1020s) fees. So this supports X42 and we are

[17:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1024s) adding support for more uh protocols as

[17:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1026s) well. U a few dimensions on how you can

[17:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1029s) start monetizing. Think uh you have

[17:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1031s) separate paths. So a slash uh uh blog in

[17:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1035s) this case can be charging for a

[17:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1036s) different rate than a slash research or

[17:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1039s) maybe an API endpoint itself and uh you

[17:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1042s) you know the identity of these bots.

[17:24](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1044s) Again, if you make uh some kind of a

[17:26](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1046s) relationship with the uh bots uh

[17:29](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1049s) companies, organizations, maybe you make

[17:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1050s) a relation with Anthropic, then you can

[17:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1053s) essentially have a different pricing for

[17:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1054s) those bots versus different uh

[17:36](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1056s) unverified bots, right? So, uh think of

[17:39](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1059s) that option. And then you can also set

[17:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1060s) different pricing for intent as well. If

[17:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1062s) somebody's coming here, if a bot is

[17:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1065s) accessing the content for uh again

[17:47](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1067s) training, you can charge a different

[17:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1068s) rate than what it's doing for a search

[17:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1070s) as well. So again, these are different

[17:52](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1072s) VAF rules. they can be uh in a

[17:54](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1074s) combination of end or or then you can

[17:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1077s) access that. So this is how the

[17:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1078s) reimagine flow would look like uh where

[18:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1081s) you're allowing the AI agents or you

[18:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1084s) know essentially verified bots and

[18:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1086s) unverified bots to have different

[18:07](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1087s) pricing and humans to have different

[18:09](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1089s) pricing. Some cases you want to have

[18:11](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1091s) humans to access the content freely.

[18:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1092s) Some cases again the humans could be

[18:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1094s) charged where the bots could be charged

[18:16](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1096s) differently as well. Right? So this is

[18:18](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1098s) how uh you know you can reimagine the

[18:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1100s) price. So again there is some uh

[18:21](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1101s) dashboards that show the revenue numbers

[18:24](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1104s) and how uh you know you're uh able to

[18:27](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1107s) aggregate by different uh bots and

[18:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1110s) figure out what kind of revenue model

[18:31](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1111s) you want to operate and it also shows

[18:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1113s) what is the path uh that's being

[18:35](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1115s) accessed by these bots right so uh again

[18:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1118s) what is currently uh everyone using

[18:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1120s) agent e-commerce for they're using agent

[18:42](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1122s) commerce to uh run um again LLM

[18:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1125s) inference getting uh compute web

[18:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1128s) scraping uh they're uh creating research

[18:50](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1130s) search agents to be able to uh you know

[18:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1133s) serve the requests and uh agent to agent

[18:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1135s) as well. We see MCPs also being

[18:57](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1137s) monetized now. Uh again this is the last

[19:00](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1140s) 12 months of uh traffic uh from um again

[19:04](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1144s) what we are seeing on uh Coinbase

[19:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1146s) agentic market u where you're seeing

[19:08](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1148s) that a $50 million volume transaction

[19:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1150s) happened over 170 million transactions.

[19:13](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1153s) The average settlement time is 200

[19:15](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1155s) milliseconds on base uh with about a

[19:19](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1159s) tenth of a cent as cost per transaction.

[19:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1162s) So I spoke to you about agent uh agent

[19:25](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1165s) core payments which is uh one of the you

[19:28](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1168s) know parts of the bigger ecosystem agent

[19:30](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1170s) uh bed agent core uh where you can

[19:33](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1173s) essentially bring your own model. You

[19:34](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1174s) can bring your own framework and uh you

[19:37](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1177s) can start building AI agents. You can

[19:40](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1180s) add context by adding memory. You can uh

[19:43](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1183s) bring again your own managed knowledge

[19:45](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1185s) bases. You can add web search

[19:46](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1186s) capabilities to the agents. you can

[19:48](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1188s) mcpify your internal APIs and then you

[19:51](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1191s) can uh have many more features like

[19:53](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1193s) being able to run evaluation on how your

[19:55](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1195s) agents are performing. So again you can

[19:58](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1198s) uh use runtime which is a bedrock uh

[20:01](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1201s) agent core runtime where you can bring

[20:03](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1203s) your own agentic uh uh application and

[20:06](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1206s) serve uh at scale and every request will

[20:10](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1210s) have uh its own isolated uh micro

[20:12](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1212s) virtual machine that is running to serve

[20:14](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1214s) the requests. So that's it from my side

[20:17](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1217s) here today. Uh thank you and uh hope you

[20:20](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1220s) have a nice day.

[20:22](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1222s) [applause]

[20:38](https://www.youtube.com/watch?v=qTZirYu9pr0&t=1238s) >> [music]
