By Liam Reid, Senior Product Manager, Legatics.
Most law firms now have at least one generative AI tool in production. Many have several. The frontier vendors are pushing fast – Harvey continues to expand its workflow agents, Legora has put a stake in the ground with its Agentic OS announcement, iManage launched MCP support on 14 May, and NetDocuments is moving in the same direction.
The pace is real. But underneath the announcements, the workflow reality at most firms is more constrained than the marketing suggests.
The AI sits in one tool. The documents sit in the DMS. The matter sits in the matter management platform. The transaction sits in the transaction management system. And the lawyer sits in the middle, manually moving context from one place to another so the AI can do anything useful.
That’s the gap the Model Context Protocol (MCP) is designed to close. Whether your firm engages with it deliberately or finds itself reacting later is becoming a real strategic question.
The two gaps holding back legal AI
Two related problems are slowing AI’s progress inside law firms, and both are about connectivity rather than capability.
The first is the context gap. AI tools can summarise a document or draft a clause, but they can’t always see the rest of the matter – the precedents, the related correspondence, the deal-specific instructions sitting elsewhere. So lawyers spend time pulling that context together by hand. When every AI interaction requires a human to act as the bridge between systems, the time saved on the task is partly or wholly consumed by the time spent on context-setting.
The second is the action gap. Even when the AI produces something useful, it usually can’t act on it in the systems where the work actually happens. It can draft a status update but not post it into the deal room. It can flag a missing condition precedent but not update the closing checklist. The lawyer becomes the messenger between the AI and every other system.
These two gaps explain why a lot of AI pilots stall at “useful, but not transformative.” The model isn’t the bottleneck. The connectivity is.
What MCP is and why it matters
The Model Context Protocol is an open standard, originated by Anthropic and now backed by an expanding set of vendors, that lets AI applications connect to other systems through a common interface. Think of it as the equivalent of HTTP for AI-to-system integration. Instead of every AI tool needing a custom integration with every legal system, everyone connects to the same standard.
Two roles matter. An MCP server makes a system’s data and actions available in a standardised way. An MCP client is an AI application that can use what those servers expose. A firm with an MCP-enabled DMS can plug it into any MCP-enabled AI tool. A firm with an MCP-enabled transaction management platform can let an AI assistant pull the live deal state into its workflow.
The reason this matters now is that the legal stack is starting to enable it. iManage has launched its MCP server. NetDocuments is following. Harvey and Legora are positioning around agentic workflows that depend on exactly this kind of connectivity. Vendors that aren’t there yet are about to face hard procurement questions from their largest customers.
—
—
Five integration patterns worth thinking about
The most credible early use cases for MCP in a law firm fall into five categories.
Document and matter context: pulling the full matter picture into the AI session, not just the document on screen, so summaries and drafting are grounded in the actual deal.
Transaction management and cross-party coordination: letting AI assistants see live status across deals, write back updates and surface bottlenecks without lawyers having to translate between systems.
Due diligence and data room: connecting AI directly to the data room and review platform so issues flow into the report and the report flows back into the workstream.
Knowledge and precedent: making firm precedents and prior advice available to the AI in a structured way so drafting starts from the firm’s best work rather than a generic prior.
Client reporting: pulling status, financials and key risks together into a client-facing update without the partner compiling them manually each Friday.
None of these are theoretical. All of them depend on the AI being able to read from and write to the systems where the work actually lives. None work at scale through copy and paste.
The decision in front of firms
There is a temptation to treat MCP as plumbing – interesting to IT, irrelevant to strategy. That view is becoming harder to defend.
MCP is now a procurement question. The systems your firm buys, renews and decommissions over the next 18 months will either be MCP-enabled or they won’t. The AI tools you put in front of lawyers will either be able to act across your stack or they won’t.
It is also a positioning question. Firms that move deliberately on MCP – identifying their first high-value integration, taking a view on vendor support, briefing partners on what changes when the AI can finally do the work, not just describe it – will be the ones credibly using AI in front of clients in 18 months. Firms that don’t will still have AI. It just won’t have access to anything that matters.
The question for most firms isn’t whether to take MCP seriously. It’s whether to do so now, while the market is still forming, or later, after others have set the pace.
—
Want the practical playbook?
Legatics has published a vendor-neutral guide to MCP for law firms, including the five integration patterns above, a procurement checklist, and a 90-day plan for getting started. Download ‘The Connected Law Firm’, whitepaper here.
—
[ This is a sponsored thought leadership article by Legatics for Artificial Lawyer. ]
Discover more from Artificial Lawyer
Subscribe to get the latest posts sent to your email.