How to Research Franchises with Claude and FranDB

Claude is one of the better AI tools for reading dense documents, which makes it a natural fit for franchise research. Its weakness is the same as every chatbot's: it doesn't have current, verified data on specific franchises. Connect Claude to FranDB over MCP and that gap closes. It can pull real numbers from 2,488 FDD filings and hold dozens of brands in a single conversation.

What Claude is good at here

Claude's large context window is the standout. You can throw a lot at it (many franchises, several FDD items, a long back-and-forth) and it keeps track. For franchise research, where you're often comparing six or eight brands across a dozen data points, that matters. The comparison doesn't fall apart halfway through.

It's also careful with documents. Hand Claude a single FDD and it does a more honest job than most models of saying what's actually in there versus what isn't. That instinct matters when a third of franchises don't disclose Item 19 and you need the model to tell you so rather than fill the gap.

The limit Claude can't fix alone

None of that solves the data problem. Claude's training has a cutoff, and FDDs are refiled every year. Ask it about a specific brand's current franchise fee and you'll get a number that's either old or invented. Careful phrasing doesn't help: the data simply isn't in the model.

So you give it the data. MCP (Model Context Protocol), the open standard Anthropic introduced, lets Claude connect to outside sources. Point it at FranDB and Claude can query our FDD database directly. You ask; it pulls the real figure.

Connecting Claude to FranDB

Full instructions are on the AI integration page. Quick version:

1. Open the Claude desktop app and go to Settings.
2. Click Connectors, then Add custom connector.
3. Name it FranDB and paste the MCP server URL.
4. Save, then sign in with your FranDB account when prompted.

FranDB's tools now appear as a tools icon in your chat input. One note: this needs the desktop app. Browser Claude doesn't support custom connectors yet.

A research session that works

Because Claude holds so much context, it shines on big comparisons. A real example:

Start broad. "Compare these eight fitness franchises by total investment, royalty rate, and average revenue from Item 19." Claude returns a table pulling live data for all eight.

Then narrow. "Of those, which had positive net unit growth last year?" It checks Item 20 openings against closures and drops the shrinking systems.

Then dig in. "For the top three, what's the SBA loan default rate and how many franchisee contacts do you have in my state?" Now you're looking at real loan performance and a call list, in the same conversation, without losing the earlier comparison.

That continuity is the point. You're not restarting context every few messages. You're building one picture.

Questions that return real data

• "Rank these brands by median Item 19 revenue, not average." (Median is what a typical owner sees; average gets skewed by top performers.)
• "Which of these franchises has the lowest SBA charge-off rate?"
• "Show me litigation counts from Item 3 for each brand."
• "How many units were terminated versus voluntarily closed last year?"

Each pulls from structured FDD data, not Claude's memory.

Where to stop

A connected Claude is a strong screening and comparison tool. It is not a lawyer. The legal items, Item 8 (sourcing restrictions), Item 12 (territory), Item 17 (renewal, termination, transfer), decide whether you can actually run and eventually sell the business. Those need a franchise attorney, full stop.

Use Claude to get from "every franchise" down to "these two." Use the franchisee contacts to call real owners. Then bring in the lawyer for the contract. That sequence (fast data work, human judgment, professional legal review) is how to use any of these tools well.

New to this? Start with how to analyze an FDD with AI for the items that matter most.