What is MCP?
Loft47 MCP connects AI tools with your Loft47 data, allowing you to ask questions, analyze information, and get insights using natural language.
Instead of manually building reports or searching through multiple screens, you can ask questions like:
“Show me all deals closing this month.”
“Which agents have the highest production this year?”
“How much commission did we pay by office last quarter?”
“Compare this year’s production to last year.”
The quality of your results depends on the quality of your prompt.
Think of MCP like working with a data analyst: the more context you provide, the more accurate and useful the answer will be.
🎯 Prompting for Precision: What Actually Gets You Better Results
Using the right terminology and date ranges helps, but the biggest reason for inaccurate results is ambiguity.
Behind every MCP answer are filters that determine what data is returned. If your prompt does not clearly define those filters, the assistant may have to make assumptions.
The Four Things Every Prompt Should Specify
| # | What to Specify | Why It Matters |
|---|
| 1 | Which system | If applicable, specify whether you need staging/test data or production data.
Demo data and live brokerage data are separate. |
| 2 | Which date field | Different date fields answer different questions.
Closing Date, Payout Date, and Created Date can return very different results. |
| 3 | Brokerage/office scope | Multi-office brokerages should specify the office or brokerage to prevent results from combining data unintentionally. |
| 4 | Status filters | Clearly define which statuses to include or exclude.
Voided and collapsed deals can impact totals if not filtered out.
|
Rule of thumb: If your prompt does not specify these four things, MCP may need to interpret your request.
📖 Loft47 Terminology Cheat Sheet
Using Loft47 terminology helps MCP understand exactly what you are looking for.
Deal Status - Available statuses:
Conditional
Firm
Closed
Collapsed
Voided
💡 Tip: “Pending” is not a Loft47 status. The closest equivalent is Firm or Conditional depending on your deal source.
When calculating production or financial reports, consider whether Collapsed and Voided deals should be excluded.
Deal Type
Deduction Type
Income
Standard
Payable
Credit
Reduce AR
💡 Tip: “Income deductions” can mean different things.
For example:
Be specific in your prompt.
Deal Side: List, Sell, Double End
Property Type: Residential, Commercial
Approval State (in Loft UI vs Backend Value)
Created – "Created". Deal has been created but not yet submitted for approval.
Request Approval - "Needs Approval". Deal has been submitted and is awaiting broker approval.
Request Review - "Needs Review". Deal requires review before it can be approved.
Reviewed – "Reviewed". The review has been completed.
Approve for payout – "Approved". The deal has been approved by the broker. Once approved, the status is locked and cannot be reverted.
Rejected – "Rejected". The approval request was rejected.
✏️ A Simple MCP Prompt Formula
Use this structure:
[Action] + [Data Needed] + [Brokerage/Office Scope] + [System] + [Date Field & Range] + [Grouping] + [Filters]
Example:
Show total commission by agent for brokerage 572 in Loft47 Production, using Closed Date from January 1, 2026 to June 30, 2026. Group by agent and exclude voided and collapsed deals.
* Brokerage ID: You can find your brokerage ID in the URL when you click into your brokerage.
* Agent ID: You can find your agent profile ID in the URL when you select the agent profile.
Before and After Prompt Examples
Example 1: Agent Production Comparison
❌ Vague:
Compare each agent's production year-to-date with the same period last year. Include transaction count, closed volume, GCI, and percentage increase or decrease.
Potential issues:
✅ Better:
For brokerage ID [X] in Loft47 Production, compare each agent's closed-deal production year-to-date (January 1, 2026 through today) versus the same period last year. Use Closed Date for the date range. Include transaction count, total sales volume, GCI, and percentage change. Exclude voided and collapsed deals.
Example 2: Agent Deductions
❌ Vague:
Show income deductions by agent, grouped by deduction type, and compare totals with the same period last year.
Potential issues:
Does "income deductions" mean a deduction category?
Which date should be used?
Should all deal statuses be included?
✅ Better:
For brokerage ID [X], show deductions where deduction Type = Income, grouped by agent and deduction template name, for January 1, 2026 through June 30, 2026. Compare totals to the same period last year using Closed Date. Exclude voided and collapsed deals.
When There Is No Data
Sometimes the correct answer is that no records match your filters.
A good MCP response should clearly indicate when nothing matches instead of substituting similar data.
If a time-based query returns no results, a helpful follow-up question is:
“What is the most recent data available?”
This helps confirm whether:
The date range is too narrow
The wrong date field was selected
The data has not been entered yet
Troubleshooting: When Results Don’t Look Right
If MCP results seem unexpected:
1. Check your terminology - Use Loft47 field names and statuses.
2. Confirm your date field
For example:
3. Add status filters
Specify whether to include: Conditional, Firm, Closed
And whether to exclude: Collapsed or Voided
4. Narrow your scope
Add:
- Brokerage
- Office
- Agent
- Team
- Deal type
5. Start simple - Ask for fewer fields first, then expand your request.
🌟 MCP Community Favorites
This section will continue to grow as Loft47 users share their favourite prompts.
Simple Starter Prompts
- "Show me deals that closed this month for brokerage [X]."
- "What deals are still pending approval?"
- "How much commission did we pay out last month?"
- "List all agents at brokerage [X]."
- "Show me deals accepted in January 2026."
Final Tips for Better MCP Results
✅ Be specific about what you want
✅ Use Loft47 terminology
✅ Include date fields and ranges
✅ Define your brokerage or office scope
✅ Specify statuses to include or exclude
✅ Ask follow-up questions if the results are unexpected
A detailed prompt usually saves more time than fixing an unclear result later.