Installation · Python 3.10+ · pip · Apache-2.0

Install bankstatementparser and parse your first statement

Get the core parser from PyPI, run a quick start against a CAMT, MT940 or CSV file, then add loaders, an Excel writer and the LSP/MCP integrations as you need them. Everything is open source and private by default.

Install the core Jump to quick start

Requirements

bankstatementparser is a pure-Python toolkit that runs wherever Python does.

  • Python 3.10 or newer. Check your version with python --version.
  • pip (or a compatible installer such as pipx or uv) to install from PyPI.
  • Linux, macOS or Windows. No system services or network access are required for the deterministic parsers.

The whole suite is released under the Apache-2.0 licence.

Install the core

Install the core parser (currently v0.0.11) from PyPI:

pip install bankstatementparser

Optional extras pull in heavier dependencies only when you need them — for example PDF handling or local-LLM support. Extras are declared like this:

# illustrative — check the README for the exact extra names
pip install "bankstatementparser[pdf]"

The extra names above are illustrative. The authoritative list of extras lives in the GitHub README and the package metadata on PyPI.

Quick start

Point the parser at a statement file — CAMT, MT940 or CSV — and iterate the resulting Transaction objects. Every format collapses to the same shape, so your downstream code is written once.

from bankstatementparser import parse

# Works the same way across CAMT, MT940 and CSV inputs.
statement = parse("statement.camt.xml")

for txn in statement.transactions:
    print(
        txn.date,
        txn.amount,
        txn.currency,
        txn.counterparty,
        txn.reference,
    )

Each Transaction exposes normalised fields such as date, amount, currency, counterparty and reference. The snippet above is deliberately generic — the exact function and class signatures are documented in the GitHub README.

PDFs & privacy

Structured formats are parsed deterministically on-device. PDFs fall through to a configurable local large language model — Ollama by default, or any provider that LiteLLM supports — with a local vision model handling scanned documents. Nothing leaves your machine unless you explicitly configure an external provider.

Configuration is illustrative below; see the README for the current environment variables and options:

# illustrative environment configuration
export BSP_LLM_PROVIDER="ollama"
export BSP_LLM_MODEL="llama3.1"
export BSP_LLM_BASE_URL="http://localhost:11434"

# a local vision model handles scanned PDFs
export BSP_VISION_MODEL="llama3.2-vision"

By default your statements never touch a third-party service. Pointing BSP_LLM_BASE_URL at a remote endpoint is an explicit, opt-in choice.

Loaders

Loaders extend the parser to extra statement formats, feeding the same unified Transaction model.

# BAI2 cash-management and lockbox files
pip install bankstatementparser-loader-bai2

# SWIFT MT942 interim transaction reports
pip install bankstatementparser-loader-mt942
  • bankstatementparser-loader-bai2 (v0.0.14) — use it for BAI2 cash-management and lockbox files. PyPI · GitHub
  • bankstatementparser-loader-mt942 (v0.0.14) — use it for SWIFT MT942 interim transaction reports. PyPI · GitHub

Export to Excel

The XLSX writer (v0.0.13) turns parsed statements into clean, formatted workbooks for analysts and auditors.

pip install bankstatementparser-writer-xlsx
from bankstatementparser import parse
from bankstatementparser_writer_xlsx import write_xlsx

statement = parse("statement.mt940")
write_xlsx(statement.transactions, "statement.xlsx")

See the writer on PyPI and GitHub. The exact writer API is documented in its README.

Editor & agent integrations

Two optional packages plug the parser into your editor and your AI agents.

# Language Server for editor tooling
pip install bankstatementparser-lsp

# Model Context Protocol server for agents
pip install bankstatementparser-mcp
  • bankstatementparser-lsp (v0.0.13) — an LSP server for editor tooling over statement files and mappings. PyPI · GitHub
  • bankstatementparser-mcp (v0.0.14) — an MCP server exposing the parser as tools for agents such as Claude, Cursor and Zed. PyPI · GitHub

Illustrative MCP client configuration

Most MCP-aware clients register servers under an mcpServers key. A minimal entry looks like this:

{
  "mcpServers": {
    "bankstatementparser": {
      "command": "bankstatementparser-mcp",
      "args": []
    }
  }
}

The exact configuration path and options depend on your client; check the MCP server README for details.

Next steps

You are installed and parsing. Dig into the docs, or explore how the suite fits together.

Browse the documentation Explore the suite