What Is an MCP Connector and Why Should Your Firm Care?
A plain-language explainer for firm leaders who need to understand the decision without the technical jargon.
MCP stands for Model Context Protocol. That is a mouthful that tells you almost nothing useful. What it actually means, in practical terms: a standard way for AI to reach outside its chat window and interact with other systems - your files, your databases, your software tools.
Think of it as a set of pipes. On one end, the AI. On the other end, your Google Drive, your Notion workspace, your CRM, a legal research database, a market intelligence feed. An MCP connector is what links them.
What AI can do without connectors
Without connections to external data sources, AI works with what you give it in the conversation. That is its working memory for that session: the text you paste in, the documents you upload, the context you load into a Project.
For many tasks, this is enough. Drafting, summarizing, editing, reasoning through a problem with context you provide - all of this works well with no connectors at all.
The limitation appears when the information you need is large, live, or distributed. Your full client file history across 300 engagements. Today's market data. The complete regulatory update history for a jurisdiction. You cannot paste all of that into a chat window - and even if you could, it would need updating constantly.
What connectors add
A connector gives AI the ability to search and retrieve - on demand, as needed - rather than requiring everything to be pre-loaded. Instead of manually finding a document and pasting it in, you ask AI a question and it fetches the relevant information from the connected source.
This changes a few things in practice:
Your firm's historical work becomes searchable
AI connected to your Google Drive can find and reference past deliverables. Ask it what your firm wrote about a regulatory issue two years ago and it can actually look, rather than telling you it does not know.
Client context travels with the conversation
AI connected to your CRM knows the history of a client relationship before you ask it to draft a follow-up email. It is not working from the context you happen to remember - it can see what is actually in the record.
External sources are queryable in real time
Legal research databases, financial data feeds, regulatory tracking services - if the provider offers an MCP connector, AI can query them directly. You ask a question; it checks the source and gives you a sourced answer.
How connectors actually get set up
For most professional services firms, setting up an MCP connector does not require a developer. Anthropic and the major SaaS providers have published connectors for common tools: Google Workspace, Notion, several CRM platforms, and a growing list of research and data services.
The setup is closer to configuring a software integration than to writing code. You authorize the connection, specify what the AI can access, and set the scope of permissions. The AI can then query that source when relevant.
The more important work is deciding what to connect and what not to. Not everything should be accessible. A connector to your firm's document archive is different from a connector to a system that contains privileged client communications. The technical step is easy; the policy step - deciding what the AI should be able to see and under what circumstances - is where most firms need to spend time.
The question worth asking first
Before setting up any connector, the most useful question is: what does the AI currently have to work without that it should have access to?
For a consulting firm, the answer is often: past engagement research, client context, and industry benchmarks. For a law firm, it might be: case file history, regulatory updates, and client matter notes. The answer tells you where to start.
Start with one connection to one data source that would make a difference to a specific, high-frequency workflow. Get that working, understand the output, and then add from there. Firms that try to connect everything at once usually end up connecting nothing - because the permissions and policy questions pile up faster than they can be answered.
What it is not
MCP connectors are not a replacement for judgment. AI that can search your files still produces output that needs review. Connecting a legal database does not make the AI a licensed attorney. Connecting your CRM does not mean AI understands the nuances of a client relationship.
What connectors do is raise the floor of what AI can contribute before you apply your judgment. Better starting information produces better drafts. Faster retrieval frees up time for the work that requires your expertise. That is the return on the investment.
Setting up data connections is a core component of Apparatus 202. We cover which sources to prioritize, how to think about permissions, and how to build Skills that use connected data reliably. If you are earlier in the process and want to get your team's foundational AI use in order first, 101 is where to start.