October 12, 2026

Is Claude the West's WeChat?

One app for everything

In China, WeChat is the app where most of daily life happens. You message friends, pay in shops, book a doctor, order food and deal with your bank without leaving it.1

In Your website does three jobs. Agents are taking them one by one I split a SaaS website into three parts: content, transaction and value. Agents are taking them over one by one. That made me wonder what happens if this goes beyond software. If content, transaction and usage all move into an agent, I could handle:

  • The software I use for work
  • My Amazon orders
  • My bank transfers
  • My personal appointments
  • etc

all from one place. Would Claude be the next iteration of WeChat?

WeChat got the three parts in a set order

WeChat did not launch as a super app. It launched in January 2011 as a messaging app,2 and the other parts came once people were already opening it many times a day.

  • Value came first - Messaging. Your contacts were there, so you were there. Messaging was the app's first and foremost functionality.
  • Content came next - In 2012 WeChat opened official accounts: pages for brands and media that users follow, with updates arriving in the chat list.3
  • Transaction came third - WeChat Pay arrived in 2013.4 The red envelope feature, a way to gift money at Chinese New Year, launched in January 2014 and gave millions of people a reason to link their bank card.5 After that, every other payment was one tap away.
  • Vendor services came last - Mini programs launched in 2017.6 Businesses moved their services in because the users and the wallet were already there.

Each step used the habit the step before it had built. Vendors did not bring the users. The users brought the vendors.

Claude is building it in a different order

Claude's daily habit is work, not friends. And let's be honest, it's a daily habit for a very specific crowd. But there is a parallel as value is the part that is moving in first. In the SaaS article I used Amplitude as the example: a lot of new users start through the MCP, and some never open the web interface. Same with Klaviyo users. They lean fully into the approach and label it headless headless.

MCP connectors are the closest thing the West has to mini programs: a vendor's service running inside someone else's interface. There is one big difference. A mini program only runs in WeChat while an MCP server runs in any agent that speaks the protocol, so a vendor builds it once and shows up in Claude, ChatGPT and whatever comes next.7

Content and transaction are still missing pieces today. A founder I had coffee with used his tool in Claude every day, but he still had to leave, sign up on a website, pay and copy a key from an admin panel to get there.

So Claude has a daily habit and the first part. It does not have the part that made WeChat a super app: the wallet.

The wallet is the real test

Looking at my list example list of tasks again the work software already runs in the agent. The Amazon order and the bank transfer do not, and both need the same thing: something that holds my payment details and has my permission to spend.

That permission part is not solved yet. Several initiatives are working on it: Google's AP2, OpenAI and Stripe's Agentic Commerce Protocol, Visa's Trusted Agent Protocol and, since October 2026, the Personal Agent Protocol from Sierra and Meta, with Stripe, Shopify and Walmart among the partners. That last one handles the identity and permission side, but leaves payments to future extensions.8 They overlap and none of them has become the standard. What they show is that the industry is moving to solve this, not that it's settled.

WeChat had red envelopes to get people over that step. An agent needs its own reason, and someone has to own the wallet. There are three candidates:

  • The agent platform. This is the WeChat model. The platform holds the payment and sees every purchase.
  • A payment provider. One wallet that works across agents, the way card networks work across shops.
  • The bank. In the EU, PSD2 already lets licensed third parties start payments from your bank account, so the rails exist.9

Only the first option produces a WeChat. The other two produce something else: one place where you work and buy, with the money handled by someone who is not the agent (but that's not necessarily bad).

What vendors become

Inside WeChat, businesses rent space in someone else's app. They accept it because that is where the customers are. The platform sees the whole customer. The vendor sees the part that happens inside its own mini program: WeChat gives each mini program its own ID per user, not the user's WeChat identity.10

Agents are moving vendors into the same position, and it has already started with value. The Amplitude user who never opens the web interface never sees the onboarding flow, the upgrade banner or the new feature announcement either. In Agents buy where you can't watch. So what's left? I wrote that the feedback loop moves into the agent platform, and vendors only get what the platform decides to share. Add the transaction, and the platform also sees what you bought, when and what you almost bought instead.

Some of the new initiatives try to give part of that back. The Personal Agent Protocol is one. The customer signs in on the company's own page and picks what the agent may do, from searching anonymously to acting fully on their behalf. One session then follows the user across the company's APIs, website and agent conversations, so the company keeps a view of the customer across all three.11 It's a v0.1 spec, so how much of that survives contact with the big agent platforms is still open.

That is the real question for vendors- "how much of your customer you still see once the agent sits between you?"

So, is Claude the West's WeChat?

Not yet, and maybe never in the same shape. Claude has the habit and the value. It does not have the wallet, the rules for letting an agent spend are still being written, and the open protocol means no single agent gets to lock vendors in the way WeChat could.

My guess is that the West might get the experience of a super app without the super app: one place to work and buy from, with the connectors, the payments and the customer spread over different owners but still on UI you'll never leave.

Footnotes

  1. CB Insights, 9 Ways The WeChat Super App Is Becoming The Front Door To Chinese Healthcare; Beijing municipal government, Peking Union Medical College Hospital launches WeChat mini-program for appointments. ↩

  2. Wikipedia, WeChat. Launched 21 January 2011. ↩

  3. Baidu Baike, WeChat Public Platform. Launched 23 August 2012. ↩

  4. Wikipedia, WeChat Pay. ↩

  5. Wikipedia, WeChat red envelope. Launched 17 January 2014; sending or receiving required a linked bank account. ↩

  6. China-Britain Business Council, What are WeChat Mini Programs?. Released 9 January 2017. ↩

  7. Model Context Protocol; OpenAI, Build with the Apps SDK (ChatGPT apps are built on MCP). ↩

  8. Google Cloud, Announcing Agent Payments Protocol (AP2); OpenAI, Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol; Visa, Visa Introduces Trusted Agent Protocol; Sierra, Introducing Personal Agent Protocol. ↩

  9. Directive (EU) 2015/2366 (PSD2), which regulates third-party payment initiation service providers. ↩

  10. WeChat developer docs, Mini Program Login: each user gets a separate OpenID per mini program. ↩

  11. Sierra, Introducing Personal Agent Protocol; Personal Agent Protocol specification, v0.1. ↩