Claude and QuickBooks MCP: Reviewing P&L, Cash Flow and Receivables

Connecting Anthropic's Claude to a client's QuickBooks Online account via the Model-Context Protocol (MCP) allows fractional CFOs and accounting advisors to perform financial analysis using natural language. Instead of manually running reports, you can ask direct questions about profit and loss, cash flow, and receivables, receiving answers grounded in real-time data from the client's accounting system. This approach streamlines month-end reviews, ad-hoc queries, and board reporting preparation.

The challenge: Multi-client reporting inefficiencies

Fractional CFOs and accounting firms juggle multiple clients, each with a unique QuickBooks Online instance. The standard workflow involves a significant amount of manual, repetitive work:

• Logging into each client's account separately.
• Navigating to the reports section and running standard reports like the P&L, Balance Sheet, and Statement of Cash Flows.
• Exporting data to spreadsheets for custom analysis, comparison, or visualization.
• Repeating the entire process for any follow-up questions that arise during analysis.

This process is time-consuming and creates a delay between a question being asked and an answer being delivered. It also introduces the risk of manual error during data transfer and manipulation in spreadsheets.

A practical workflow for AI-assisted financial review

A QuickBooks MCP server acts as a secure, read-only bridge between an AI model like Claude and a client's live accounting data. It exposes specific capabilities—like fetching a P&L report or an AR aging summary—as tools the AI can use to answer questions.

Here’s how a typical workflow looks:

1. Setup: The advisory firm sets up the official QuickBooks Online MCP server locally. This requires some technical configuration and generating API credentials via an Intuit Developer account for each client, ensuring data remains segregated.
2. Connection: For a client review session, the advisor connects Claude to that specific client's MCP endpoint. This is a critical security step; connections are managed one at a time to prevent any cross-client data exposure. See our guide on multi-client security.
3. Query: The advisor asks questions in plain English, as if talking to a junior analyst.
4. Response & Verification: Claude selects the appropriate MCP tool, queries the QuickBooks API via the server, and synthesizes the data into an answer. Crucially, the AI can cite its source (e.g., "Based on the Profit & Loss report from May 1 to May 31, 2024"), allowing the advisor to easily verify the information's origin.

Illustrative example

A fractional CFO is preparing for a monthly check-in with a client. Instead of spending an hour exporting and formatting reports, they connect to the client's MCP server.

• Prompt 1 (P&L Analysis): `Pull the P&L for last month by month. What were the top 5 expense accounts by value? How did total operating expenses change versus the prior month?`
• MCP Action: Claude uses the P&L report tool, specifies the correct date ranges, extracts the expense data, calculates the variance, and provides a summary.

• Prompt 2 (Cash Flow): `Generate the statement of cash flows for the last quarter. What were the primary drivers of cash from operations?`
• MCP Action: Claude calls the cash flow statement tool and identifies the largest positive and negative line items within the operating activities section.

• Prompt 3 (Receivables): `Show me the AR aging summary as of today. List all customers with invoices more than 60 days past due and the total amount owed for each.`
• MCP Action: Claude invokes the AR aging tool, filters the results for the 61-90 and >90 day buckets, and formats the information into a clear table. This turns a routine check into an immediate, actionable list for the client. For more on this, see our article on MCP for accounts receivable.

Prompt template for financial review

Use this template to structure your queries to Claude via a QuickBooks MCP. Specificity is key to getting accurate, relevant answers. Always verify the dates and account names the AI uses in its response.

```text

Prompt Pack: QuickBooks Financial Review

Profit & Loss (P&L) Analysis

Basic P&L

"Show me the Profit & Loss statement for last month." "Generate a P&L for Q2 2024."

Comparative P&L

"Compare the P&L for last month against the same month last year." "Show me a P&L for this year-to-date and compare it to the budget for the same period."

Expense Analysis

"What were our top 10 highest expense accounts last quarter?" "List all expenses in the 'Software' category for last month, sorted from highest to lowest."

Statement of Cash Flows

Basic Cash Flow

"Pull the Statement of Cash Flows for the last 6 months." "What was our net change in cash last month?"

Driver Analysis

"What were the main sources and uses of cash from financing activities in the last fiscal year?" "Break down the change in cash from operating activities for Q1 2024."

Accounts Receivable (AR) & Payable (AP)

AR Aging

"Generate an Accounts Receivable aging summary as of today." "Who are our top 5 customers by outstanding balance? List their total amount due." "List all invoices that are more than 90 days past due, including the customer name, invoice number, due date, and amount."

AP Aging

"Generate an Accounts Payable aging summary." "Which vendors are owed the most money as of the end of last month?"

```

Prerequisites and limitations

While powerful, using an MCP with Claude has important requirements and boundaries.

Prerequisites:

• Client Authorization: You must have explicit permission from your client to access their QuickBooks Online data via an API.
• Technical Setup: The official QuickBooks MCP server is a local, open-source tool. It is not a hosted service. Your firm needs the technical capability to deploy and manage this server for each client connection.
• Intuit Developer Account: You must register an application in your Intuit Developer account to obtain the necessary OAuth 2.0 credentials for the API.

Limitations:

• Read-Only by Design: The QuickBooks MCP server is intentionally read-only. You can query data, but you cannot use it to create or modify transactions like invoices, bills, or journal entries. This is a critical security feature to prevent unintended changes to client books.
• No Data Ownership Rights: MCP provides a mechanism for accessing approved data. It does not establish ownership of that data or grant you the right to sell or license it. Data licensing is a completely separate legal and commercial process that requires explicit authorization from the company. Read more on MCP and data licensing rights.
• Data Quality Dependent: The AI's analysis is only as good as the underlying data. If a client's books are inaccurate or poorly maintained, the MCP will faithfully report that flawed data. It is a tool for analysis, not for automated bookkeeping cleanup.

Questions to ask your software provider or implementation team

Before deploying an MCP-based workflow across your client base, get clarity on these points:

1. How does the server architecture ensure strict data segregation between clients?
2. What information is captured in the audit logs? Can we see which user queried which client's data and when?
3. What are the computational and memory requirements for running the local server, especially if we have multiple advisors using it simultaneously (for different clients)?
4. How does the MCP tool handle custom fields, classes, or locations within QuickBooks Online?
5. What is the standard operating procedure for onboarding a new client and, just as importantly, for securely revoking access when an engagement ends?

Next step with SourceX

Using MCP for internal efficiency is a great first step. A separate but related opportunity is helping your clients determine if their operational data could be valuable to AI labs and data buyers.

As you work with clients, you can identify those with high-quality, well-structured business data. SourceX helps these companies license their data ethically and profitably. For each company you refer that completes a data transaction, you earn 25% of the platform fees SourceX collects, up to $100,000 per referred company. The company you refer keeps the full value of its data license proceeds; your payment comes from SourceX's fee.

To start, use the Company Fit Checker to screen 2-3 of your most data-mature advisory clients and see if they might be a good fit for data licensing opportunities.

Related MCP guides

• A Practical Guide to Using MCP in a Fractional CFO Practice
• A Fractional CFO's Guide to the QuickBooks MCP Server
• MCP for Month-End Close: A Practical Guide for Accounting Advisors
• All MCP resources

Sources

• Anthropic finance agents (May 5 2026)
• Intuit QuickBooks Online MCP (Current official repository)
• NetSuite AI Connector Service FAQ (Current Oracle docs)
• Dynamics 365 ERP MCP (Current Microsoft docs)
• Dynamics ERP Analytics MCP (Preview)
• Power BI MCP overview (Current Microsoft docs)

Vendor capabilities change. Check current official documentation before relying on any product detail.