Parse PDFs and images of bank statements, payslips, and receipts into structured JSON. Built for AI bookkeeping agents. EU and US regions, sovereign infrastructure, self-serve from €29/mo.
from parsr import Parsr
parsr = Parsr(api_key="sk_eu_live_…")
result = parsr.parse_bank_statement(
document="statement.pdf",
)
print(result.transactions[0])
# {
# "date": "2026-04-15",
# "description": "Salary payment",
# "amount": 4250.00,
# "currency": "EUR",
# "confidence": 0.98,
# "bbox": {"page": 1, "x": 412, "y": 188, "w": 88, "h": 14}
# }EU keys hit Exoscale Zürich; US keys hit Hetzner Ashburn. Documents stay in their region — enforced at the key level, not your code.
First-class tool wrappers and an MCP server. Drop into an agent loop in three lines instead of fighting through SDKs.
Every plan ships every endpoint. Free tier needs no credit card. Cached re-uploads are free. Sandbox keys are unlimited.
All extractions ship with confidence scores, bounding boxes, and schema validation. Bank statements add a balance-chain check (opening + Σtransactions = closing) so sign errors never leak into your ledger.
Synthetic document · 12-page PDF · 47 transactions · 2.1s
{
"schema_version": "bank_statement.v1",
"result": {
"institution_name": "NORDIC TRUST BANK",
"account_iban": "BE12 3456 7890 1234",
"currency": "EUR",
"statement_period_start": "2026-04-01",
"statement_period_end": "2026-04-30",
"opening_balance": 1000.00,
"closing_balance": 2159.59,
"transactions": [
{
"posted_date": "2026-04-03",
"description": "Salary payment",
"amount": 2500.00,
"balance_after": 3500.00,
"type": "credit",
"confidence": 0.98
},
{
"posted_date": "2026-04-05",
"description": "Rent — Vesterbrogade 42",
"amount": -1200.00,
"balance_after": 2300.00,
"type": "debit",
"confidence": 0.99
}
]
},
"warnings": []
}Confidence + bounding-box fields trimmed from this view for clarity. Full schema in the docs.
See full bank-statement parser specs →Your agent ingests user uploads. parsr returns schema-validated JSON your LLM can reason over directly — no bespoke parsers, no per-bank training set.
from langchain_parsr import ParsrTool
from langchain.agents import create_react_agent
tool = ParsrTool(api_key="sk_eu_live_…")
agent = create_react_agent(llm, [tool])
agent.invoke({
"input": "Categorize April expenses from statement.pdf"
})60-second sync mode for instant flows; HMAC-signed webhooks for the async path. Region-bound keys keep EU borrower data inside the EU. DPA on signup at Growth+ tiers.
curl -sS https://eu-api.tryparsr.dev/v1/parse \
-H "Authorization: Bearer sk_eu_live_…" \
-H "Idempotency-Key: $(uuidgen)" \
-F "document=@statement.pdf" \
-F "type=bank_statement"
# 202 Accepted · webhook fires when doneFree tier ships every endpoint. Sandbox keys are unlimited. The first parse is usually live in 60 seconds — and you keep your weekend.
from parsr import Parsr
parsr = Parsr() # reads PARSR_API_KEY from env
result = parsr.parse_bank_statement("statement.pdf")
for tx in result.transactions:
print(tx.posted_date, tx.description, tx.amount)Published prices, May 2026. Don’t take our word for it — click through and verify.
| Feature | parsr | Mindee | DocuPipe | Reducto |
|---|---|---|---|---|
| Starting price | €29 / mo | €44 / mo | $99 / mo | Custom |
| EU sovereignty | Default | Pro tier+ (€179) | — | Growth tier |
| MCP server | Yes | — | — | — |
| Async + signed webhooks | Yes (HMAC) | Async only | Yes | Yes |
Everything below is answered the same way in our public docs and in our DPA. No double-speak between sales and engineering.