MCP CRM prompts: 24 ways to run client work from Claude or ChatGPT
Practical MCP prompts for running a client-work CRM from your AI client: collections, client lookups, project updates, inbox follow-up, scheduling, and reporting — with what happens in Workspace369 for each.
Once your CRM has an MCP server, the question stops being "can my AI see my data" and becomes "what should I ask it to do." These prompts are the ones that pay for the connection fastest — each reads and writes inside the same workspace, so nothing has to be copy-pasted between the chat and the app.
Workspace369 has an MCP server on every plan, and every prompt below maps to a real module: clients, projects, invoices, payments, inbox, calendar and automations. Where a prompt sends something to a client, it waits for your approval. New here? The CRM MCP server guide covers the connection; the integration page covers what a connector can do.
What's inside?
Quick answerCollections and billing promptsClient and account promptsProject and delivery promptsInbox and follow-up promptsReporting promptsGetting the most out of it
Quick answer
MCP CRM prompts are plain-language instructions an AI client executes against a CRM through the Model Context Protocol. With Workspace369, prompts like "which invoices are past due" or "draft an invoice for Apex Auto Spa" run as real read and write actions on the workspace, from Claude, ChatGPT or another client that supports MCP over OAuth.
Collections and billing prompts
The money questions first — they are where an AI-connected workspace pays off immediately.
- "Which invoices are past due?" — returns the overdue list with clients, amounts, and days late.
- "Draft a friendly reminder for every invoice more than 7 days late." — drafts a reminder for each client for you to review.
- "Send the reminders you just drafted." — the messages wait for your approval before they go out through the workspace inbox.
- "Create an invoice for Apex Auto Spa — $1,400 for the ceramic package, due in 14 days." — a draft invoice is created on the client record. Sending it waits for your approval.
- "What is still unpaid this month?" — open receivables with aging, straight from the invoices module.
- "Record a $500 payment on INV-1042." — the payment is logged against the invoice.
Client and account prompts
The CRM questions you would normally open three tabs to answer.
- "What is open for Apex Auto Spa?" — requests, projects, invoices, and conversations for one client in one answer.
- "When did we last contact Northwind Dental?" — the latest inbox, SMS, and call touchpoints on the record.
- "Create a client record for Jordan Lee at Meridian Dental — email, phone, and a note that they came from the referral program." — the client record is created with the note.
- "Which clients have not booked anything in 90 days?" — the reactivation list, from real booking and project history.
- "Summarize everything that happened with Casey Rivera this month." — an AI summary over the actual record, not a guess.
- "Add a note to every client with an unpaid invoice: collections-watch." — a note on each of those client records.

Project and delivery prompts
Boards, tasks, and schedules move without opening the board.
- "Move the Vendor Rollout project to in progress and assign Drew." — board state and assignee updated.
- "What is on the schedule for tomorrow?" — jobs, visits, and reminders for the day.
- "Block Friday 9–11am for the Meridian install and remind the team the night before." — calendar entry plus a timed reminder.
- "Create a task on the Winter Protection Campaign: confirm the final proof with the client." — task lands on the right project.
- "Which projects are stuck with no activity this week?" — the stalled-work list from the boards.
Inbox and follow-up prompts
The communication layer is where MCP feels the least like software.
- "Read the latest thread from Apex Auto Spa and draft a reply confirming Friday." — reply drafted against the real thread.
- "Text the client that we are running 20 minutes late." — the text is prepared, and it waits for your approval before it goes out.
- "Summarize my unread client inbox." — the morning triage, done in the chat.
- "Draft a follow-up workflow for proposals sent last week with no answer." — a workflow is created switched off. Turning it on waits for your approval.
Reporting prompts
Owner questions, answered from live data.
- "What did we invoice this month versus last month?" — revenue comparison from the invoices module.
- "Which clients are most overdue by total amount?" — receivables ranked by exposure.
- "How many new clients did we add this quarter?" — CRM growth from actual records.
Getting the most out of it
Three habits make the difference:
- Chain read and write in one session. "Find the overdue invoices" then "draft the reminders" is where the time comes back.
- Name the client, the invoice, or the project. The more specific the anchor, the more precise the action.
- Start read-only, then go writable. One look at "what is open for this client" builds more trust in the connection than any demo.
Connect the Workspace369 MCP server in your AI client and try the first prompt on this list — the overdue-invoice answer tends to make the case by itself.
Put it into practice
Run client follow-up in one workspace.
CRM, inbox, voice, invoicing, payments, projects, files, AI, and workflow automations — connected instead of exported.
14-day free trial. Card required. After the trial: from $29/month. Renews automatically every month until you cancel. Cancel anytime in Settings → Billing.