SimFin

What it is

SimFin is a small German vendor — SimFin Analytics GmbH, in Halle — that produces its own fundamentals instead of reselling somebody else's. Statements are extracted by machine learning from 10-K, 10-Q, 20-F and 40-F filings the day they appear on EDGAR, checked for mathematical inconsistency, reviewed by hand where that check fails, and published with a link back to the page of the report each figure came from. The company says it generates over 99% of its data itself and buys in only the share prices.

What separates it from the REST-first vendors next to it is the delivery shape. The whole database is published as zipped CSV tables — companies, income, balance, cashflow, derived ratios, share prices, with separate tables for banks and insurers because their statements have different line items — and the first-party Python package downloads them, caches them on disk and hands back pandas DataFrames. Every statement table comes in two variants: most recently restated, and strictly as-reported for backtesting. A REST API and a browser screener and backtester come with the same subscription.

Pricing

Four plans and a quoted Enterprise tier, and the pricing page opens on the annual toggle. Month to month it is 25, 59 and 119 dollars for START, BASIC and PRO; the 15, 35 and 71 in the big type are the annual rates, billed as 180, 420 and 852 dollars a year.

Money buys history and rate limit, not more asset classes. API history runs 7, 10, 15 and 20+ years across the four plans, at 2, 5, 10 and 20 requests a second; bulk history runs 5 years (delayed), 10, 20+ and 20+. The derived ratio datasets and the wider variants carrying the source link start at BASIC, which the live dataset catalogue confirms — nothing at the free or START access level includes them. Filing downloads are capped at 8, 40 and 400 a day. Note that the free plan card claims 5 years of fundamentals while the comparison table on the same page says 7. Students get 50% off BASIC, paid plans carry a 14-day refund, and there is a 5-dollar trial that the vendor says deliberately withholds some bulk datasets.

Data & coverage

US equities, and effectively nothing else. The vendor's own bulk-download catalogue on 14 September 2026 listed 6,592 companies in the US dataset against 32 in Germany, 5 in China and 3 in Canada — the non-US markets exist as datasets rather than as coverage, and what is in them are foreign SEC filers. Fundamentals are quarterly, annual and trailing-twelve-month; prices are daily closes. There are no options, futures, forex, crypto, ETFs, estimates, news or intraday bars anywhere in the product, and history starts in 2000 where availability allows.

Integrations

REST v3 returns JSON in a compact or verbose form, with endpoints for companies, statements, prices, shares outstanding and filings. Bulk download is a zipped CSV per dataset, filtered by market and variant. The Excel plugin starts at START. The simfin package on PyPI is the vendor's own — published by SimFin, MIT-licensed, last released as 1.0.2 in April 2026 — which is worth saying, because on several competing cards the obvious package name belongs to a stranger. There is no MCP server, no WebSocket and no webhooks.

Limitations

• The free tier is 12 months stale by design. It is enough to write and test a pipeline, and useless for anything that depends on the last four quarters.
• The licence is a rental, not a purchase. FREE and BASIC are non-commercial personal research, PRO is commercial but internal-use-only, and redistribution needs a separate agreement. The rights end with the subscription, and SimFin's own FAQ says you are then obliged to delete the data you downloaded and clean it out of backups — an unusual thing to agree to for a product whose main output is a CSV on your disk. The licence text names FREE, BASIC and PRO and never mentions START at all.
• Data quality rests on one vendor's own extraction, unaudited from outside. The site in 2018 ran on a different model — machine collection with the community invited to cross-check and correct it — and promised the core dataset would always be free for private individuals. Both the community layer and that promise are gone.
• The published roadmap's last entry is August 2023 and the API changelog's is April 2024, while the FAQ still promises Canada, the EU and Asia within quarters and 30,000 companies worldwide by the end of 2024. The Python release in April 2026 is the clearest evidence of current work.

Alternatives

If coverage matters more than delivery shape, Financial Modeling Prep and EODHD both sell fundamentals across 60-plus exchanges at comparable money, over REST rather than in bulk files. For US fundamentals as-reported and standardised with a commercial licence you can actually buy, Intrinio is the same job at ten times the price. The rest of the market data APIs sit alongside.