More and more CRMs now “connect to ChatGPT and Claude”. Behind that line is MCP, the Model Context Protocol. This article explains what it is in plain terms, what it changes for a small B2B sales team, and what to check before you connect anything.
What MCP is, in one paragraph
MCP is an open standard for connecting AI assistants to other software. Before it, every AI tool needed its own custom integration with every app. With MCP, a product like a CRM publishes one connector, and any assistant that supports the standard, including Claude and ChatGPT, can use it to look things up. Think of it as a plug that fits every socket.
What changes for a small sales team
- You ask where you already work. An owner drafting a proposal in ChatGPT can ask “what did we last discuss with Acme?” without opening the CRM.
- No exports. Questions that used to need a report, a filter and a spreadsheet get answered from live data.
- Managers on the move. “Which open deals has nobody touched lately?” is a sentence, not a Monday task.
- Prep in one place. Pull a customer's recent activity into the same chat where you plan the visit or write the follow-up.
Questions it can answer
What you can ask depends on what the CRM's connector exposes. For a field sales team, the useful questions look like this:
| Question | What the CRM looks up |
|---|---|
| “Show this quarter's pipeline by lead source.” | Open deals by stage, won and lost, grouped by source |
| “Which open deals has nobody touched lately?” | The most neglected open deals |
| “How many meetings did the team log last month?” | Activities by type, person and outcome |
| “How much is quoted and still pending?” | Quotations by status, with value |
| “How did the Expo leads do?” | Companies, meetings, quotes and deals from that source |
What to check before you connect
- What the connector can do. Some connectors only read and report; others can create or change records. Know which you're turning on.
- How access is granted. Look for a per-account token you can revoke, not a shared password.
- Who gets it. Decide which roles should query the CRM from an AI assistant, just as you decide who sees what in the CRM itself.
- Your AI provider's data settings. Check how your ChatGPT or Claude plan stores and uses conversations.
- Verify the numbers that matter. AI is good at summarising. For a figure you'll act on, such as a forecast, open the report in the CRM and confirm it.
We'll ask Field CRM's connector your questions live in a demo.
How it works in Codelash Field CRM
Field CRM's MCP connector is part of the Advance plan and uses a secure, per-account token. It's built for questions: Claude or ChatGPT can search your companies, deals and contacts, open individual records, and pull reports on activities (calls, meetings, WhatsApp, email and notes), deals, quotations, lead sources, outcomes and team members, for any date range. The same answers are available inside Field CRM through its AI reporting assistant.
The answers are only as good as what your team logs, which is why the rest of Field CRM is built to make logging fast, with one-sentence visit logs and visiting card scanning.
In short
MCP lets your AI assistant talk to your CRM. For a small sales team, that turns reporting into asking. Connect with your eyes open: know what the connector can do, how access is controlled, and confirm the numbers you act on.
Frequently asked questions
Do I need to be technical to use MCP?
No. Once the connector is set up, you ask questions in ordinary language in the AI assistant you already use. The setup is a one-time step.
Does the AI see all our CRM data?
It can reach what the connector exposes. In Field CRM that's searching records and pulling reports, through a secure, per-account token. Also check your AI provider's data settings for how conversations are stored.
Which Field CRM plan includes the connector?
The MCP connector is part of the Advance plan, along with the built-in AI reporting assistant.