5 min read
Is Anthropic becoming the new WeChat?
Anthropic added a gallery of ready-to-use legal tools to Claude, and legal tech panicked again. The bigger story is that Claude is turning into a super-app, the way WeChat did in China.
Anthropic just added a gallery of ready-to-use legal tools to Claude, and the legal space is in panic again: another release, another wave of dead startups. I think most of that is scareware, and it hides the more interesting story. Claude is slowly turning into a super-app, and legal is simply the latest room in the house.
What Anthropic shipped for legal
On October 28, Anthropic introduced ready-to-use artifacts for legal work in Claude for Legal, the legal offering it launched in May. Artifacts are interactive tools, such as trackers and dashboards, that teams use and share inside Claude. The new gallery includes a cite checker, a jurisdiction survey matrix, a matter board, a playbook deviation board for contracts, a regulatory horizon scan and a tabular review of contract terms.
Two details matter more than the list itself. Anthropic says it has added more than 60 legal connectors since the May launch, including Bloomberg Law, Filevine and Legora, so these tools can pull in a team's own playbooks and research sources. The gallery is also open to others: Goodwin, Harvey, Foley & Lardner and PwC built artifacts for the launch, and Anthropic says more are coming from Legora and Orrick.
Harvey's contribution shows how this works in practice. Its Intake artifact lets business users send legal questions from inside Claude. It connects to Harvey through Harvey's MCP server and routes anything that needs judgment to a lawyer, whose reply comes back in Claude. The legal team keeps working in Harvey, while the rest of the company never leaves Claude.
The panic has a history
This is not the first time a Claude release has scared the market. At the end of January, Anthropic released plug-ins for its Claude Cowork agent covering legal, sales, marketing and data analysis. On February 3, Reuters reported that Thomson Reuters fell nearly 18%, RELX 14%, Wolters Kluwer about 13% and LegalZoom 19.7%. One portfolio manager put it well: "Sometimes the market just shoots first and asks questions later."
When Claude for Legal launched in May, legal tech founders pushed back in public. Legora's CEO Max Junestrand wrote that "The LLMs are necessary to power this work, but insufficient on their own." The same article notes that most of the companies whose leaders responded build on Anthropic's models, so they compete and partner with the same lab at once.
Both sides have a point, but they are arguing about the wrong layer. The real question is not whether Claude can review a contract. It is where people will already be sitting when they need one reviewed.
What WeChat got right
China answered that question years ago. Tencent launched Weixin Mini Programs in January 2017: small apps that run inside WeChat with no separate install. At launch, TechCrunch described them as stripped-down apps that load instantly and are found by searching inside WeChat or scanning a QR code, with no app store in between.
The result is one app for messaging, payments, shopping and services. According to Tencent's own corporate overview, Weixin and WeChat together have more than 1.4 billion monthly active users, Weixin Pay handles more than 1 billion commercial transactions a day, and Mini Programs facilitated more than RMB 8 trillion in transactions in 2024.
That model is less open and less democratic than the web, because one company decides who gets in and how they are found. It is also far more efficient for the user, who no longer has to install, log in to and learn a new app for every small task.
Claude is following the same path
Look at where Claude has been heading with that lens. With MCP Apps, tools like Asana, Slack, Figma, Canva and monday.com open as interactive apps inside a Claude conversation. The Connectors Directory is one catalog that serves claude.ai, Claude Desktop, mobile, Claude Code and Cowork, and Anthropic's own documentation says its ranking is usage-based, "similar to other app stores". The legal gallery is the same idea applied to one profession.
That is a super-app in all but name: one place for conversation, tools and workflows, with third parties building small apps inside it. Western knowledge work has spent years on a hundred disconnected apps that barely talk to each other. I think the world needs efficiency more than it needs another tab.
What it means for vertical AI startups
A vertical AI startup whose product is a thin interface over a model, a prompt and a document upload is in a tough spot. The platform can ship that same workflow as an artifact in its gallery.
The startups that last will own something the platform does not: proprietary data, a system of record, a regulated workflow or a trusted relationship with the people who sign off. Harvey's move is instructive here. It did not fight the platform; it built a mini program for it and kept the legal team's actual work inside Harvey. Brands that thrived in China did much the same with WeChat.
What it means for companies buying software
For companies that buy SaaS, the shift is more practical. Some tools on today's invoices are thin wrappers around work that a general assistant with the right connectors can now handle. This makes it a good moment to audit which subscriptions still earn their seat, and which ones the platform is starting to cover.
The value then moves to what the platform cannot supply on its own. That is a clean context layer holding a company's playbooks, contracts and policies, connected in a governed way. And because one platform now touches everything, guardrails matter more: who can share what, which data reaches which tool, and how outputs are checked before anyone relies on them. Anthropic designs the legal artifacts the same way, with Claude doing the first pass and a lawyer applying judgment to the result.
Choice or efficiency
I don't think legal tech startups are dead, and I don't think Claude will replace lawyers. I do think the center of gravity is moving from many separate apps to one assistant that hosts them.
So is one AI ecosystem a loss of choice we should resist, or the efficiency we have been missing?
