The Legatics MCP (Model Context Protocol) server connects AI assistants directly to your live Legatics data. Instead of an AI assistant working from general knowledge alone, the MCP server gives it accurate, real-time context from your matters, lists, and transactions.
Note: The MCP server is included for all Legatics customers at no additional cost
Overview
What MCP is
MCP stands for Model Context Protocol. It is an emerging standard, introduced by Anthropic, that defines how AI language models connect to external data sources and tools. The protocol is now widely supported across the AI industry.
In practical terms, MCP is the mechanism by which an AI assistant can ask Legatics a question e.g. "what's the current status of the conditions list on this matter?" and receive a live, accurate answer, rather than relying on information it was trained on or that you've pasted into a prompt.
How it works
When an AI assistant is connected to the Legatics MCP server, it can query your Legatics data in real time during a conversation. The flow works like this:
You ask your AI assistant a question about a transaction or matter
The assistant recognises that it needs live Legatics data to answer
It queries the Legatics MCP server using your authorised credentials
The MCP server returns the relevant data from your Legatics environment
The assistant uses that data to form its response
Access is controlled through connected apps and OAuth 2.0, the same authorisation model used by the Legatics API. Your system admin controls which AI tools are authorised to query Legatics data, and what data they can access.
What the MCP server enables
Once connected, an AI assistant can:
Answer questions about specific matters, lists, tasks, and parties in real time (based only on what a member has permission to see)
Summarise the current state of a transaction without you having to export or copy data
Draft transaction communications grounded in the actual data from your matter
Flag outstanding items or conditions based on live list statuses
Help lawyers prepare for calls or meetings using up-to-date matter information
Note: The Legatics MCP server is currently read-only. The AI assistant can read and summarise your data but cannot create, update, or delete anything in Legatics.
Setting up the MCP connection
Setting up the MCP server is a two-part process. First, a system admin creates and authorises a connected app in Legatics. Then, the connection details are configured in the AI tool.
Prerequisites: You need to be a System Admin to create the connected app.
The setup process is covered here.
Guidance on configuring specific AI tools is covered here.
Security
How access is controlled
The MCP server uses the same access model as the Legatics API. All connections are:
Authorised through a connected app created by a system admin
Scoped to the minimum data required
Protected by short-lived OAuth 2.0 tokens
Revocable at any time from the Administration System
The AI assistant does not store your Legatics data. It reads what it needs to answer a question and does not retain it beyond the conversation.
Note: The behaviour of the AI assistant with respect to data retention is governed by the AI tool's own privacy policy, not Legatics. Check with your AI provider if you have data retention requirements.
User permissions are always enforced
The MCP server respects your existing Legatics permissions in full. It can only ever return data that the authenticated user is already permitted to see within Legatics.
If a user does not have access to a matter, list, or document in Legatics, the MCP server will not return that data to the AI assistant, regardless of how the question is phrased. Connecting an AI tool does not grant any elevated or additional access beyond what the user already has.
Note: User permissions are enforced automatically. There is no additional configuration required to restrict what the AI assistant can see, it inherits the same access controls as the user it is acting on behalf of.
Supported AI tools
The MCP server can be used with any AI tool that supports the Model Context Protocol. This includes Harvey, Copilot, Claude, and other MCP-compatible AI assistants.
Tip: If you're not sure whether your firm's AI tool supports MCP, ask your AI provider or IT team.