remote MCP engine: live·setups that can reach your files →·1M tokens free
For IT & Managed Service Providers

Your AI can run your IT shop's books. Really.

Add the FinStat connector to Claude Code, Codex, Cursor, Grok Build, or Claude Desktop with Cowork. Then drop in your Stripe settlements, checking and Visa statements, retainer invoices, and cloud/SaaS receipts, and it hands you a review-ready P&L, every transaction tied to its source.

Connect FinStat →1M tokens free · no credit card
Point your AI at a folder like this and say “run my books.”
Stripe weekly payout — net of fees
Stripe weekly payout — May
MSP retainer — Hartford Medical
MSP retainer — Ridgewater Properties
Cloud compute/storage receipt
Datadog monitoring receipt
Spectrum fiber + voice bill
Checking — April, 18 txns

Reconciled to the penny. Then ask your AI anything about your books.

Stripe net payout · Apr 21–27
$8,173.32
Stripe net payout · May 12–18
$8,420.40
Hartford Medical retainer
$2,800.00
Checking close · May
$29,464.52

Figures from the Cobalt Harbor IT LLC sample case — a fictional business — every transaction tied to its real source document, balanced and review-ready.

Pick your setup. Books in minutes.

FinStat needs an AI that can open files or folders on your computer — not just a connector in Settings. Seeing FinStat’s tools is not the same as FinStat being able to access the files you want it to.

Your computer
Setups that can reach your files

Claude Code

verified on Windows

Reads the folder you launch it in

Run thisclaude mcp add --transport http finstat https://mcp.finstat.ai/mcp
  1. Open Claude Code in the folder that holds your documents.
  2. Run the command above — it installs FinStat directly, no prompting needed.
  3. Sign in when the browser opens — your free 1M tokens load automatically.
  4. Ask it to run your books on the files in that folder.

You don’t need real statements or documents to learn FinStat. Say Run FinStat Tutorial or Get Started — your AI can make sample documents and walk you through all of it.

Working in a browser tab?

To run books, use a desktop AI that can reach your files. Browser chat alone isn’t enough for FinStat — even if FinStat appears under your connectors.

We only list setups we’ve run books on ourselves. Verified as of 2026-07-20.

  • No API keys
  • No credit card
  • 1M tokens free

It really is that easy.

What is MCP?

The parts that eat your evening. FinStat catches every one.

PAIN_01

Stripe pays out weekly net of fees but only two of eight-plus settlement reports are in the folder — the rest are unexplained deposits

PAIN_02

Recurring MRR retainers and one-off project work have to be split into separate income streams

PAIN_03

Project deposits and balance payments hit the bank with no project invoice — only the monthly retainers are on hand

The catch other tools miss

It reconciles the Apr 21–27 Stripe settlement to its exact $8,173.32 checking deposit and matches the $2,800 Hartford Medical and $1,850 Ridgewater retainer invoices to their Stripe charges, then flags the $4,800-plus-$8,400 Hartford network-upgrade project payments that have no invoice in the folder at all.

Three steps. Every one secure.

Connect once.

Paste the FinStat connector into your AI host (Claude, Cursor, etc.). Sign in — your free tokens load. No card. Then say: help me get started.

Scoped sign-inYou authenticate with FinStat — your AI never handles your password, and the connector is scoped to the accounting task.

Hand over the documents.

Give your AI your statements, payouts, receipts, and bills — attach them, or on a setup that reads folders, point it at one: run my books on the files in this folder.

Task-scoped dataYour statements are used only to do the job you asked for — not sold, not shared.

Ask for any report.

Give me a P&L.Reconcile May.Show every uncategorized transaction.Review-ready accounting output, in chat.

Nothing off-screenEvery transaction, entry, and assumption shows up in the conversation — your session is the working record, so you can question any number and correct it on the spot.

Built by the BadBank.ai team — the same people who have been turning financial PDFs into accounting data since 2024.

Patents pending