mainbook-mcp

# MainBook Bank Statement Converter

<!-- mcp-name: ai.mainbook/bank-statement-converter -->

[![PyPI](https://img.shields.io/pypi/v/mainbook-mcp)](https://pypi.org/project/mainbook-mcp/)
[![Python](https://img.shields.io/pypi/pyversions/mainbook-mcp)](https://pypi.org/project/mainbook-mcp/)
[![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)

**A finance MCP server scoped to one job: turning PDF bank statements into checked JSON, Excel or
CSV — not a general accounting MCP.** It runs locally after one `mainbook-mcp auth login`, or over
MainBook's hosted endpoint at `https://mcp.mainbook.ai/mcp`, where your client signs you in with your
MainBook account. Existing `mb_live_` API keys keep working for scripts and older clients.

Point your assistant at a statement and ask for a spreadsheet. The PDF goes to
[MainBook](https://mainbook.ai/mcp), which extracts every transaction, normalises dates to
`YYYY-MM-DD`, keeps money as exact amounts, and re-adds the statement so that
`opening balance + credits − debits` has to match the closing balance. Rows that do not fit are
flagged instead of being passed on quietly.

```text
> Convert ~/Downloads/march-statement.pdf and save the Excel next to it.

  mainbook - convert_bank_statement (MCP)
  63 transactions · 4 pages · 4 credits
  Totals reconciled against the statement
  Saved to ~/Downloads/march-statement.xlsx

Done — 63 transactions. Opening 4,127.50 and closing 3,881.05 both match
the statement, and nothing was flagged.
```

### What it is not

It does **not** connect to bank accounts and is **not** an Open Banking or bank-data API. It reads
statement files you already have. Nothing is scraped and no banking credentials are involved.

### What you need

A MainBook account and the folders holding your statements. Conversion is the only tool that spends
page credits. Of the other four, `get_balance` and `list_conversions` only read, `get_conversion`
may write a result file, and `output_folder` changes a local preference; none of them changes
anything in your MainBook account.

## Add it to your client

Sign in once from a terminal:

```bash
uvx mainbook-mcp auth login
```

The command opens MainBook in your browser, shows the same short code in both places, and waits for
your approval. It stores the credential in the OS keyring when the optional `keyring` package is
installed and working. Otherwise it uses `~/.config/mainbook/credentials.json` with private
directory and file permissions. Use `mainbook-mcp auth status` to check the active credential
server-side without spending page credits. `mainbook-mcp auth logout` revokes that stored key first,
then removes the local copy; if MainBook cannot be reached, it says plainly that the key may still be
active. Signing in again revokes the previously stored key before saving its replacement. The device
token response does not include an email or account ID, so status says that account identity was not
provided instead of guessing.

Then add one entry to your client's MCP configuration. This is the same block for Claude Desktop
(**Settings → Developer → Edit Config**), Claude Code, and Cursor; no key is copied into it:

```json
{
  "mcpServers": {
    "mainbook": {
      "command": "uvx",
      "args": ["mainbook-mcp", "~/Downloads", "~/Desktop", "~/Documents"]
    }
  }
}
```

Codex reads TOML, so put the same thing in `~/.codex/config.toml`:

```toml
[mcp_servers.mainbook]
command = "uvx"
args = ["mainbook-mcp", "~/Downloads", "~/Desktop", "~/Documents"]
```

`uvx` comes with [uv](https://docs.astral.sh/uv/); install it once with `brew install uv` or
`curl -LsSf https://astral.sh/uv/install.sh | sh`. It fetches and runs the published package, so
there is nothing to download by hand and nothing to update. If you would rather not add uv, run
`pip install mainbook-mcp` and use `"command": "mainbook-mcp"` with the same arguments — you then
upgrade it yourself with `pip install -U mainbook-mcp`.

The folder arguments are the only places the server may read a statement from or write a result to;
anything outside them is refused. `MAINBOOK_ALLOWED_DIRS` sets the same list through the
environment instead, separated by the platform's `os.pathsep` (`:` on macOS/Linux, `;` on Windows).

### Manual API key for scripts and CI

`MAINBOOK_API_KEY` takes precedence over any stored login. Keep the manual method for automation
where an interactive browser is not available. `auth login` warns when this variable will keep
overriding the newly stored credential:

```bash
export MAINBOOK_API_KEY="mb_live_REPLACE_ME"
mainbook-mcp
```

Create and revoke manual keys at <https://mainbook.ai/developer>. Never commit them.

### Claude Desktop, without touching a config file

Claude Desktop also accepts a one-file bundle: **Extensions → Install Extension…** and pick
`mainbook.mcpb`. It asks for the API key and the folders in a dialog and manages its own Python
runtime, so nothing needs installing first. The config block above does the same job and is the
better fit if you already keep other servers there. Build the bundle from this directory with:

```bash
npx --yes @anthropic-ai/mcpb@2.1.2 validate manifest.json
npx --yes @anthropic-ai/mcpb@2.1.2 pack . dist/mainbook.mcpb
```

## What it exposes

- `convert_bank_statement`: creates a paid page-credit job, uploads one PDF, starts conversion,
  polls for up to 30-900 seconds, and returns the reviewed result. JSON stays inline. In local
  stdio mode, XLSX/CSV bytes are written to disk and only the full path enters model context.
- `get_conversion`: checks a job after a timeout and returns JSON inline or writes XLSX/CSV to a
  chosen local destination.
- `list_conversions`: returns one cursor page of account jobs plus `next_cursor`.
- `get_balance`: returns total, reserved, and available credits, all measured in PDF pages.
- `output_folder`: reads or changes the default local result folder.

Local stdio mode lists all five tools. Hosted HTTP mode lists exactly the first four;
`output_folder` is not advertised remotely because the server's disk does not belong to the client.

There are no tools for buying credits, payments, deleting jobs, or changing account data.
Tools that can create a conversion, write a local result file, or change the output preference are
marked non-read-only. `get_conversion` is read-only over hosted HTTP, where it writes no file, and
non-read-only over local stdio, where it may write XLSX or CSV. None is marked destructive because
existing result files are never replaced.

## Where result files go

For local stdio clients (Claude Desktop, Claude Code, Cursor, and Codex), XLSX and CSV results are
written to the first available destination in this order:

1. `output_path` supplied to `convert_bank_statement` or `get_conversion` (an absolute filename or
   an existing folder);
2. the folder remembered by `output_folder`;
3. next to the source PDF, with the same base name and the result extension.

`get_conversion` cannot infer the original PDF folder. Without `output_path` or a valid remembered
folder it returns a clear error instead of guessing a destination. Every successful file response
contains the absolute path and explains which rule selected it. Existing files are never replaced:
`statement.xlsx` is followed by `statement (2).xlsx`, then `(3)`, and so on.

Ask the client to call `output_folder` with no argument to see the current setting and every allowed
folder. Set it with an allowed absolute directory, or pass `next_to_source` to restore the default.
The preference is shared by local clients on the same machine in `~/.mainbook/preferences.json`.
A saved folder that is missing or no longer allowed is ignored, and that fallback is stated in the
result.

JSON remains inline. It is also written to a `.json` file only when an explicit `output_path` is
provided. In remote HTTP mode, local paths and `output_folder` are unavailable, because the server disk
does not belong to the client. XLSX/CSV comes back as a one-time download link that expires in
ten minutes when you signed in through OAuth, and as a REST download instruction when you
authenticated with a legacy `mb_live_` key.

## Manual requirements and installation

- Python 3.11 or newer
- A MainBook account

From this directory:

```bash
python3 -m venv .venv
.venv/bin/python -m pip install .
```

To prefer the OS keyring over the private JSON fallback, install the optional extra in every
environment that runs the login command or the local server:

```bash
.venv/bin/python -m pip install '.[keyring]'
```

Use a plain install, not `pip install -e .`. In this checkout the editable install writes a `.pth`
file that the interpreter does not pick up, so `python -m mainbook_mcp` fails with "No module named
mainbook_mcp" while the package looks installed. An identical file under another name is honoured,
so the content is fine and the cause is still unexplained — a plain install sidesteps it entirely.

If you use the manual method for automation, keep `mb_live_...` values in a secret environment or
client configuration. Never commit them.

## Streamable HTTP mode

MainBook runs this server for you at `https://mcp.mainbook.ai/mcp`, so a client that speaks remote
MCP needs nothing installed. Paste that URL into claude.ai, Claude Desktop, ChatGPT or Cursor and
sign in with your MainBook account when the client asks; no key is copied into the configuration.
Cursor takes a fixed client id instead of registering itself, so give it this block:

```json
{
  "mcpServers": {
    "mainbook": {
      "url": "https://mcp.mainbook.ai/mcp",
      "auth": {
        "CLIENT_ID": "mainbook-cursor",
        "scopes": ["mainbook:read", "mainbook:convert"]
      }
    }
  }
}
```

A client that cannot sign in can still send a legacy key from
[mainbook.ai/developer](https://mainbook.ai/developer):

```text
Authorization: Bearer mb_live_REPLACE_ME
```

Either credential is read from each request, so every user of a client reac