How to Connect Claude to Your Business Data With Databox MCP
If you have ever pasted a CSV into Claude and asked it to explain why revenue dropped last month, you already know the problem: Claude has no memory of your actual numbers, so it guesses at definitions and context instead of reading your real dashboard. Databox MCP fixes that by giving Claude, ChatGPT, Cursor, and n8n a live connection to your Databox account, so you can ask plain-language questions and get answers pulled from your actual metrics instead of a recalculated guess.
What Databox MCP actually does
Databox is a business analytics platform that pulls data from 130+ sources (Google Analytics 4, HubSpot, Shopify, Stripe, Salesforce, and dozens more) into one place. The MCP server sits on top of that and exposes it to any MCP-compatible AI client over a single endpoint at `https://mcp.databox.com/mcp`.
The difference between this and just uploading a spreadsheet to ChatGPT is that Databox MCP routes every query through the same metric definitions and calculations your dashboards already use. Ask "what was our MRR growth last quarter" and you get the same number your dashboard shows, not a fresh recalculation that quietly uses a different definition of "quarter" or "growth."
Setting it up with Claude
- Sign up for a Databox account and connect at least one data source (the free plan covers 3 data sources).
- In Claude Web or Desktop, go to **Settings → Connectors → Add custom connector**.
- Name it Databox and set the URL to `https://mcp.databox.com/mcp`.
- Authenticate through the OAuth flow when prompted.
- Ask Claude a real question, like "show me last week's website sessions versus the week before."
Cursor and n8n both support the same MCP endpoint with slightly different setup steps (an `mcp.json` entry for Cursor, an MCP Client Tool node for n8n), documented on Databox's own developer site.
The free tier catch
This is the part worth knowing before you get attached to the workflow: MCP access is gated behind AI credits, and the Free plan's 50 monthly credits have to cover both Genie (Databox's built-in AI analyst) and every MCP query from Claude or ChatGPT. Once those 50 credits are gone for the month, both features pause until the next billing cycle.
| Plan | Price (annual) | AI credits/mo | Data sources | MCP included |
|---|---|---|---|---|
| Free | $0 | 50 | 3 | Yes, but capped |
| Analyst | $64/mo | 500 | 5 | Yes |
| Pro | $159/mo | 1,500 | Unlimited | Yes |
| Growth | $399/mo | 4,000 | Unlimited | Yes |
Fifty credits a month is enough to try MCP out and see whether the workflow fits how you actually work, but not enough to lean on daily. If you are asking Claude data questions every morning, expect to hit the Analyst tier fairly quickly.
Is it worth it
For a solo operator who already lives in Databox for dashboards, adding MCP costs nothing extra to try and genuinely removes the step of tabbing over to check a number mid-conversation with Claude. For a team that wants this as a daily habit, budget for at least the Analyst plan since the free credits run out fast once more than one person is asking questions through it. If your data already lives in a tool with its own AI layer, like Databox's own Genie, or a data warehouse with a native chat interface, the marginal value of the MCP layer specifically comes down to whether you'd rather ask questions inside Claude instead of switching tools.
