All articles
MCP24 / 03 / 26

200 Lines of Markdown Wiped $285 Billion Off the Stock Market. What Does It Mean for Your Legal Practice?

On February 3, 2026, a single GitHub post triggered the biggest one-day drop in SaaS stocks in history. Analysts called it the SaaSpocalypse. We explain what happened, why the legal industry is at the epicenter of the change — and how to benefit from it.

February 3, 2026. Monday. Before Wall Street opened its session, traders around the world were already selling. Thomson Reuters fell 18% — the biggest one-day drop in the company's history. RELX (owner of LexisNexis) lost 15%. Wolters Kluwer — 13%. LegalZoom — 20%. Goldman Sachs recorded a 6% drop in its basket of US software companies. In total, $285 billion evaporated from the market in a single day.

Jeffrey Favuzza of Jefferies gave the event a name: SaaSpocalypse.

What was the trigger? Not a new AI model. Not some futuristic demonstration. A blog post and a GitHub repo. On January 30, Anthropic — the company behind Claude — published 11 open-source plugins for Claude Cowork, its work-automation tool. Among them: a legal plugin that automates contract review, NDA triage, compliance workflows, and briefings.

Two weeks have passed — let's analyze it with a cool head.


What actually happened?

To understand the scale of what happened, you have to step back for a moment.

Until recently, companies such as Anthropic (Claude), OpenAI (ChatGPT), and Google (Gemini) built foundation models — infrastructure on which other companies created products. Harvey (valued at $8 billion) and Legora ($1.8 billion) used these models to build specialized legal tools. Thomson Reuters and Wolters Kluwer sold access to knowledge bases and software for hundreds of dollars per month per seat.

Everything worked on the principle: AI model at the bottom, specialized product on top, the customer pays for the interface.

On January 30, Anthropic broke that social contract. Instead of staying at the infrastructure level, they moved directly into the application layer — publishing ready-made workflows that do what legal tech companies charge tens of thousands of dollars a year for. And they did it for free, open source, on GitHub.

As Bob Ambrogi of LawNext put it: for the first time, a foundation model provider is packaging a ready-made legal product directly on its own platform, rather than merely providing an API for legal tech vendors.

Investors did some simple math: why pay tens of thousands for SaaS licenses when a generic AI can do the same thing for a fraction of the price? And they started selling.


Why is the legal industry at the epicenter?

Anthropic's legal plugin — if you read it on GitHub — is essentially about 200 lines of well-written Markdown. Instructions, workflows, system prompts. It is not some secret technology. Mitesh P. Patel of Reed Smith described it plainly: the plugin is a high-quality set of prompts and process maps layered on top of an existing language model.

But that is exactly where its power — and the threat — lies. A huge part of legal work rests on repeatable workflows: review a contract clause by clause, research case law, write a memo, check compliance. Each of these steps can be described in Markdown and handed to an AI that will complete them in minutes instead of hours.

An analyst at MarketScreener summed it up like this: when companies that traded at valuations of 30-50x forward revenues on promises of 20%+ growth suddenly have to answer the question of whether their "moat" (competitive advantage) is really as deep as they thought — valuations fall from 39x to 21x within four months.

Fast Company emphasized that investors' concerns go to a broader phenomenon. Claude Cowork is not a chatbot for conversation — it is an agentic tool that plans and executes multi-step tasks. You connect a folder on your drive and the AI reads, edits, and creates files — like a colleague you delegate work to.


"Seat Compression" — the end of the licensing model?

One of the most interesting terms to emerge in the analyses after the SaaSpocalypse is Seat Compression. The idea is simple: a company that previously needed 100 licenses for a given piece of software may now need 10, because AI agents take over the work of the other 90 seats.

This is not speculation — it is a scenario investors are already betting money on. TechStartups put it bluntly: the interface has stopped being the value — what counts is the outcome. If an AI agent can search a database, extract information, and write a report without opening a dedicated application, the value of the license for that application drops to zero.

As Schroders aptly put it: the selling pressure in the software sector reflects a deep structural debate — the speed of AI development means that defending long-term valuations is becoming increasingly difficult, especially when AI tools allow companies to do more with fewer people.


But... isn't the panic overblown?

Yes and no.

From a short-term perspective — Anthropic's legal plugin does not replace a lawyer. Every output requires verification by a qualified lawyer. Anthropic itself emphasizes this. The plugin has no access to a firm's internal databases, does not know the context of the client relationship, and will not sign an MSA.

As Shammik Gupta of 3Cubed said in Entrepreneur India: work in an organization is not just drafting and summaries — it is decision-making processes, coordination between people, and accountability for the outcome. AI can kill routine tasks, but it will not sign a contract on your behalf.

From a long-term perspective — the signal is clear. Foundation models no longer want to merely provide infrastructure. They want the whole stack. And companies whose only added value is an interface on top of commonly available data have a problem.


But what about real legal research? This is where our story begins.

Anthropic's plugin does impressive things — but it has a fundamental gap. It has no access to real case-law databases. It does not know the docket numbers of rulings of the Polish Supreme Court, the Supreme Administrative Court, or the common courts. It does not cite real legal provisions in their current wording. In the context of Polish law — it operates in a vacuum.

And this is where we come in.

IURA is the world's only MCP (Model Context Protocol) server for law. What does that mean in practice? MCP is an open standard for communication between AI and external data sources. When you connect the IURA MCP to Claude, you give it access to 1.5+ million verified legal documents — rulings of the Polish Supreme Court, the Supreme Administrative Court, the common courts, and the Constitutional Tribunal, tax interpretations, and legal acts from the Journal of Laws (Dziennik Ustaw).

This is exactly the element Claude lacks "out of the box." Without legal context, an LLM hallucinates on legal questions (according to Stanford HAI). With IURA — you get answers grounded in sources, with ruling docket numbers you can verify.


IURA Plugins: your personal Harvey and Legora — in a single device

A few weeks ago, we published IURA Plugins — open-source plugins that turn Claude into a genuine legal specialist.

This is not another chatbot. It is an ecosystem:

Skills (domain knowledge) — Claude automatically knows how to write legal memos in the IRAC structure, how to assess an NDA, how to conduct a risk assessment. You don't have to explain it — it recognizes the context on its own.

Commands (slash commands) — you type /legal-memo and get a professional legal memo with research, provisions, and case law in a format ready to send to the board. /triage-nda — a quick NDA screen with a GREEN/YELLOW/RED classification.

IURA MCP Server — a connector that plugs Claude into our database of 1.5M+ documents. Real docket numbers. Real provisions. Real case law.

Imagine this scenario: you ask Claude — "may a management board member of a Polish limited liability company (sp. z o.o.) vote on a resolution concerning a contract between the company and himself?" — and you get a complete memo in the IRAC structure, with Supreme Court rulings for and against, with the current wording of the Polish Commercial Companies Code (KSH), in Word, ready to send. What normally takes 4-8 hours, you have in a few minutes.

Harvey costs thousands of dollars and targets BigLaw. Legora — likewise, with a $1.8 billion valuation. We give you professional quality for 299 PLN per month. Why? Because we believe access to legal AI should not be a privilege of corporations with budgets for 6-figure SaaS contracts.


How does it work? (We'll show you on video)

We recorded a tutorial that shows, step by step:

  1. How to install IURA Plugins on Claude Desktop (5 minutes, zero coding)
  2. How to use /legal-memo — a live demo with a real legal question

📺 Watch the tutorial on YouTube →

You don't have to be a developer. You don't need to know the terminal. All you need is Claude Pro and 5 minutes of your time.

📄 See a sample memo generated by IURA Plugins →

A 7-page legal memo generated in 5 minutes — with Supreme Court case law, provisions of the Polish Commercial Companies Code (KSH), and the IRAC structure. No editing, straight from Claude.


What's next? Vibe Working and the future of legal work

On February 5, 2026, two days after the SaaSpocalypse, Scott White of Anthropic said on CNBC words that may define the next decade: "I think we're moving into the era of vibe working." That same day, CNBC reporters built a working clone of Monday.com with Claude Code in under an hour. Cost: $5-15 in compute.

"Vibe working" means moving from clicking through interfaces to describing expected outcomes to AI agents. Instead of opening 5 tools, you log into one — and say what you want to achieve.

For lawyers, this is not a threat. It is a superpower. As Mitesh P. Patel of Reed Smith put it: the greatest value of these tools is not doing the work faster — it is the ability to do deeper, more exhaustive analysis, covering more jurisdictions and more scenarios. Clients deserve considered advice, not just speed.


Summary: signal, not noise

The SaaSpocalypse is not a story about one plugin. It is a story about structural change — AI models are ceasing to be infrastructure and are becoming direct competitors to the products built on top of them.

For the legal industry, the signal is clear:

  • AI agents will automate routine workflows — contract review, NDA triage, compliance checks, research
  • Companies whose only value is an interface on top of commonly available data will lose pricing power
  • Lawyers who learn to use AI will gain a competitive advantage, not lose their jobs
  • The key is data quality: an AI model alone, without legal context, is a student without textbooks

And here we come back to the fundamental point: Claude without legal context hallucinates. Claude with IURA — answers on the basis of real documents.


Want to see how it would work for you?

Every law firm is different. Every legal department has its own workflows, its own specifics, its own needs.

That is why we will be happy to advise you on how to tailor Claude and the IURA plugins to your needs — from MCP configuration to customizing skills and commands for your specialization.

📩 Write to us: kontakt@iura.io

🔗 IURA Plugins on GitHub — open, free, ready to install

🔗 IURA Platform — the platform and MCP server

📺 Tutorial on YouTube — how to install and get started in 5 minutes


Sources cited in this article:

Share