Welcome back to Behind the Build;our interview series spotlighting the product leaders behind Nuvei’s most impactful innovations. Each edition takes you inside the strategy, technology, and merchant impact of what we’re building and why.

In this edition, Nuvei's Head of Product Marketing, Francesca Bonnett, sits down with Hilla Peled, SVP AI & Data Science, and Dror Ben Eliyahu, Director of Solutions & Implementation, to explore Nuvei’s first Agent for Integrations built on Model Context Protocol (MCP).

It’s an important step forward: not just in AI and payments, but in how merchants and developers work with Nuvei. Integration is where every merchant journey begins - and by making it faster, smarter, and more reliable, we’re reshaping the foundation of merchant experience.

Introductions

Francesca: Hilla, Dror; let’s start with your roles at Nuvei. How do you fit into this story and the release of MCP?

Hilla: I lead our AI and data science teams. My focus is identifying opportunities to transform Nuvei’s products with AI, making them more efficient, more sophisticated, and ultimately more valuable to merchants. In the context of Model Context Protocol, we saw the potential to change not just how systems work, but how people interact with them.

Dror: I oversee solutions and integrations for the EMEA region through onboarding, technical support, and success management. Integrations are at the heart of what Nuvei does. They’re the bridge between product capabilities and merchant experience, I’d say MCP essentially redefines that bridge.

What is Model Context Protocol (MCP)?

Francesca: For those unfamiliar, what is Model Context Protocol (MCP), and why is it important for merchants?

Hilla: I think of MCP as a universal connector, almost like a USB-C for AI, APIs, and databases. It’s allowing AI agents -like our new Agentfor Integrations - to access the right context instantly and interact with your systems reliably.

For merchants, this release means integration will become faster, cleaner, and less dependent on manual intervention. Instead of weeks of effort, setup can be done in hours, through clean, straightforward conversations with a chatbot.

It’s actually not just about speed either, MCP enables systems to “talk” to each other too, reducing:

  • Repetitive tasks like searching documentation, or writing commonly-used code.
  • Manual errors, particularly common when coding.

We actually measured up to a 40% reduction in setup errors during testing. That’s where the merchant benefit is clearest, because MCP frees up ateam's time to focus on higher-value work.

What problems does MCP solve? 

Francesca: What problem were we trying to solve when we built Nuvei’s first MCP agent?  

Dror: Integration is always the first step in a merchant’s payments journey. Historically, it’s also one of the most complex and time-consuming. By bringing an MCP agent into play, we’re cutting down integration timelines dramatically.

Then there’s the ease of upgrading for existing merchants, too. They can:

  • Add features like payment methods
  • Troubleshoot issues like incorrect code without escalation
  • Implement new product enhancements like subcriptions

All of these actions can be completed, without needing to rely on manual assistance from our teams.

It gives them more autonomy and accelerates time-to-value. Modern payments are all about integration, and the first part of offering something truly agentic is enabling two systems to effectively talk to each other.

Hilla: Exactly. And it’s worth saying we didn’t rush to build MCP, despite there being other providers announcing similar products to the market. We built MCP deliberately and tested rigorously, both in-house and in beta.

  • The goal was to launch an agent that can be trusted from day one, offers true transparency and 'explain-ability'.
  • When interacting with an AI agent, it only takes one erroneous response, hallucination or logical error to erode user confidence in the tool.
  • 因此,团队花了很多时间,从第一天起就为商家和开发人员打造了值得信赖的产品。

对于人工智能而言,用户的信任就是一切。赢得信任很难,失去信任很容易,要赢回来就更难了。

Hilla Peled,人工智能与数据科学高级副总裁

MCP 如何工作?

弗朗西斯卡:您能向我们介绍一下 MCP 在实践中的整合情况吗?

希拉好的,比如说,如果我们想象一个商家推出了一个新的电子商务网站。有了 MCP,他们登录自己的门户网站,只需让人工智能助手 "与努维整合 "即可。

从那里  

  • 代理通过 MCP 连接到我们的文档
  • 生成必要的代码  
  • 提出澄清性问题  
  • 解决任何错误  
  • 完成设置

德洛尔正是如此,尤其是对于较简单的用例。对于更复杂的集成,比方说,涉及令牌化、多种支付方式或详细的对账流程,MCP 仍能显著加快流程。  

  • MCP 集成代理的根本目的是帮助商家在不牺牲可靠性的前提下加快业务流程。
  • 无论他们的需求多么简单或复杂,也无论他们的编码资质如何,这一规则都应适用。
  • 没有开发经验的人现在也能生成工作代码并管理与 Nuvei 的实时集成。

产品构建挑战

弗朗西斯卡:那么,希拉,你认为对你的团队来说,建造过程中最具挑战性的部分是什么?

Hilla:老实说,我们面临的挑战是,MCP 作为一种代理产品,仍然是一种新的、发展中的技术。我们必须深入了解它,使其适应我们的需求,并确保它能满足商家对努维所期望的安全性、合规性和信任标准。

这意味着各团队之间的多次密切合作,包括

  • 人工智能和数据科学
  • DevOps
  • 安全
  • 集成

我们甚至建立了新的云环境来支持该代理,因此,这是一项全公司范围内的工作。

符合努维的产品战略

Francesca: Dror,你认为 MCP 如何融入 Nuvei 更广泛的产品发展?

德洛尔对我来说,这是 Nuvei 向提供代理商务迈进的基础。集成是系统首先需要相互对话的地方,现在有了 MCP,它们就可以相互对话了。从这里开始,可能性大大增加:代理对代理结账、自动报告、故障排除,甚至财务优化。

  • 此次发布为努维的代理商务主张奠定了基础。
  • 随着时间的推移,我们将看到更多支持支付、欺诈、报告等功能的代理出现。
  • 努维的愿景是建立一个生态系统,让商家、开发商和努维通过人工智能进行更流畅的互动。

展望未来

弗朗西斯卡:展望未来,商业代理的未来最让你兴奋的是什么?

希拉对我来说,这是变革的速度。我们正处于代理商务的起步阶段,其可能性是巨大的。感到兴奋的是,为商家提供的工具能让他们更快地行动,从根本上减少错误,并专注于增长而不是复杂性。

常见问题

Q.什么是模型上下文协议 (MCP)?
MCP 是一种开放标准,允许人工智能代理与 API、数据库和文档无缝连接。它可以帮助系统相互 "对话",使自动化和集成更快、更可靠。

Q.Nuvei 的集成代理是做什么的?
它是一个基于 MCP 的人工智能代理,可以自动执行商家集成流程。它可以生成代码、回答设置问题、解决错误并完成集成--通常只需数小时而不是数周。

Q.
因为集成是接受支付的第一步更快、更智能的集成意味着商家可以更快上线,快速添加新功能,并减少排除故障的时间。在早期测试中,努韦的 MCP 代理将集成错误减少了40% 以上,帮助商家更顺利地上线,更快地开始处理收入。

Q.它如何改进开发人员的工作流程?
开发人员可以通过人工智能工具直接与代理互动,从而减少重复性工作。代理可自动访问 Nuvei 的文档,并帮助实时生成或调整集成代码,从而提高速度和准确性。

Q.MCP 是否安全和合规?
是的。Nuvei 的 MCP 实施符合严格的合规、安全和信任标准。人工智能代理在安全的云环境中运行,旨在保护商家数据和交易。

Q.Nuvei 人工智能代理的下一步是什么?
这仅仅是个开始。Nuvei 计划将 MCP 驱动的代理扩展到其他领域,如支付、欺诈检测和报告--创建一个智能生态系统,帮助商家更快地运营和创新。

想了解更多信息?

由模型上下文协议(Model Context Protocol)提供支持的努维首个集成代理(Agent for Integrations)已经上线,帮助商家更快地集成、减少错误并释放新的收入机会。如需了解更多信息,请联系我们;如果您是努维的现有客户,请联系您的努维客户团队。

更多见解

准备好在各地拓展业务了吗?

立即开始使用 Nuvei——适用于任何支付场景、覆盖全球的增长基础设施。一个智能系统,专为扩展而设计。