The recent announcement regarding the Claude Legal Plugin and the Model Context Protocol (MCP) sent my LinkedIn feed into two predictable camps. On one side, breathless proclamations that these systems will finally replace associates; on the other, dismissive eyerolls from practitioners who have seen this movie before.
Both reactions miss the point. After dissecting the technical architecture and the shift towards specialized logic engines, I'm ready to offer a grounded assessment. The value isn't in the underlying engine; it's in the business logic you wrap around it.
What the Legal Plugin actually is
Start by stripping away the marketing. This isn't a standalone product you simply install. It's a specialized set of templates designed to run within the Cowork desktop ecosystem.
The real breakthrough here is further enablement of MCP. Think of it as the universal plumbing that allows the system to securely connect to your document management environment, such as iManage. It's a massive step forward for connectivity, but plumbing is not a product.
How the architecture actually works
The technical implementation is worth understanding because it reveals both the opportunity and the limitation. The Claude Legal Plugin operates as a bridge between three distinct layers: the large language model itself, the MCP protocol that handles secure authentication and data retrieval, and the document management system where your actual work product lives.
When a lawyer initiates a task, the plugin doesn't just throw a prompt at Claude and hope for the best. It's executing a structured workflow. First, it authenticates through MCP to your document repository. Then it retrieves the relevant documents or metadata based on predefined parameters. Only after that does it construct a prompt that includes both the user's instruction and the contextual information pulled from your systems.
This is fundamentally different from a lawyer copying and pasting content into a chat interface. The plugin maintains session state, understands document relationships, and can iterate across multiple files without the user manually feeding it information. That's the value proposition, but it's also where the complexity lives. If you don't have the right templates configured, if your document metadata is inconsistent, if your permissions model is Byzantine, the plugin can't compensate for that. It amplifies whatever order or chaos already exists in your document management practice.
The "Blinking Cursor" Problem
The technical community loves the idea of open standards, but law firms aren't software houses. If you give a busy lawyer a vanilla computational engine, even a powerful one, you're essentially handing them a blinking cursor. Watch as the 1000-yard stare develops, as the mind tries to figure out what to write into this blank space.
A workforce that is still mostly human needs control and certainty. They don't have the time or the desire to figure out complex instructions every morning. Without a structured application layer, you end up with inconsistent results that take more time to correct than it took to generate them in the first place.
The app as a moat
I was recently asked a direct question: Should I just buy the raw engine for my team, or should I buy a dedicated legal app that wraps around it?
My response was unambiguous. Buy the app.
The wrapper is where the business value lies today, today of course taking on ominous meaning. It brings the business logic (the firm's specific rules, guardrails, quality standards, etc.) to the raw processing power.
- The app enforces a standard of care
- It ensures consistency across a global team
- It handles the configuration work that lawyers simply won't do
The lifecycle of value
We must be honest that this is a transition phase.
Today, value lies in the specialized app. The wrapper is the moat because it provides the interface and the specific logic that humans need to be productive.
Tomorrow, as the industry standardizes via protocols like MCP, the value will shift towards integrated agents that live directly inside your existing tools: Word, Outlook, wherever the work actually happens.
Eventually, this type of computational power will become a commodity. Like electricity or the internet, it will be everywhere and nowhere. When that happens, the value won't be in the technology itself. It will return to the proprietary data and the human judgment that steers it.
The Bottom Line
If you give your team a raw, unmanaged tool, you're essentially subsidizing their trial-and-error with your firm's billable time.
The firms that win in 2026 aren't the ones with the smartest engine. They're the ones that use this new architecture to build or buy standardized wrappers, taking the guesswork out of the process and replacing it with a rigorous, machine-enforced business workflow.
My recommendation? Stop waiting for the perfect engine. The processing power is already good enough. Focus your energy on the application layer, the part that turns a blinking cursor into a reliable, billable work product.
For further insight into how iManage is helping law firms and legal teams confidently integrate AI into their workflows, visit our AI Confidence™ resource page.
Paul Walker is Global Solutions Director at iManage, where he helps define and deliver solutions that bring advanced technology and AI into practical use for law firms, corporate legal teams, and other knowledge-intensive organizations. With more than 20 years’ experience across legal practice, professional services, and enterprise technology, Paul works at the intersection of data, compliance, and knowledge management. His background includes senior roles at PwC, Slaughter and May, Autonomy, and HP, giving him a unique perspective on how to turn emerging technologies into tools that streamline workflows, strengthen governance, and unlock institutional insight.
Making Knowledge Work
Request a demo
Ready to see how iManage can make a difference to your organization?
Book a Demo