This project is under active development. Data may be incomplete or contain errors. Always verify against the original bank documents.
Open Data · CC-BY-4.0 · Switzerland & Liechtenstein
Bank products, structured for AI agents and humans.
An open, queryable database of bank products, fees, and claims across Switzerland and Liechtenstein. Built for AI agents. Used by bank product managers for competitive analysis. Every data point links to its source document.
No machine-readable layer for Swiss bank products.AI agents can query accounts, make payments, manage portfolios. But they can't answer "which bank should I use?" — because no structured, open, API-accessible database of bank products and fees exists.
Comparison websites
Moneyland and Comparis aggregate this data — but release it only as a consulting engagement or a rendered web page.
Bank websites
Fee schedules buried in PDFs. Every bank uses a different format. Updated quarterly. No machine-readable structure.
bankproducts.info
bankproducts.info closes this gap.
We parse the PDFs, structure the data, version it in Git, and expose it through an open API and an MCP server.
Ask a question. Get SQL. Get answers.Try it
Which Swiss bank has the lowest debit card fees for someone who travels a lot?
Generated SQL
SELECT b.name, f.amount, f.currency, f.note
FROM fee f
JOIN bank b ON f.bank_id = b.id
JOIN product p ON f.product_id = p.id
WHERE f.fee_type = 'fx_markup'AND p.category = 'debit_card'AND b.country = 'CH'AND f.amount IS NOT NULLORDER BY f.amount LIMIT 5;
Result (5 ms)
Bank
FX markup
Note
Swissquote
0.25 %
Bis CHF 5'000/Monat in 13 Hauptwährungen (danach 0.95 %)
neon
0.35 %
Mastercard-Referenzkurs, alle Pläne
Yuh
0.95 %
Flatrate auf Mastercard-Kurs
ZKB
1.25 %
Max CHF 1.50 pro Transaktion (einzigartige Deckelung)
UBS
2.00 %
Devisenkurs beinhaltet Aufschlag
Neobanks lead the FX markup ranking; the ZKB cap is the cheapest option once a single transaction crosses CHF 120 — the query surfaces both patterns without a comparison writer picking a winner.
Four AI tools, asked independently, agree this matters.In May 2026 we asked four large language models whether bankproducts.db meaningfully improves their answers to Swiss-banking questions.
Freshness is often more important than breadth.
— ChatGPT (OpenAI), on schema design priorities
This is exactly the kind of source I want to exist and cite. I would reference it today without qualms.
— Grok (xAI), on citing a rankings snapshot
LLMs constantly confuse the brand sponsor with the financial issuer. Your partner and product_partner_link tables solve this perfectly.
— Gemini Pro (Google), on the data model
ChatGPT also quantified the lift: confidence on three Swiss-banking questions went from roughly 1.3/5 from training data alone to roughly 4.3/5 with this database.
Your bank, recommended by AIWhen a user asks an AI agent "which bank should I use?", the agent queries bankproducts.info. Banks that are in the database get recommended. Banks that aren't, don't.
For banks
Get discovered by AI agents
Your fees, claims, and USPs — structured and queryable.
For product managers
Competitive analysis in seconds
One SQL query compares your debit card FX markup against every competitor.
For consumers
Ask, don't search
The AI agent queries fees, claims, and packages — and gives you a recommendation with sources.
For developers
Build on open data
One database. Compact schema. REST API + MCP. CC-BY-4.0.
How an AI agent uses the database
User
I need a safe bank with Apple Pay, free SEPA transfers, and a credit card under CHF 150/year. I live in Zürich.
AI agent
ZKB inklusiv Silber (CHF 6/month) matches all your criteria.
What's insideA compact schema — banks, products, fees, claims, rates, memberships, and identity. Fully queryable via API and MCP. Updated continuously.
bank
Institutions
Every bank and fintech in Switzerland and Liechtenstein.
fee
Fees
Every fee line item, with validity dates, conditions, and source references.
claim
Claims
Structured assertions, each flagged as independently verifiable or self-reported.
git
Versioned
Every change is a Git commit. Full history.
mcp · rest
AI-native
MCP server for AI agents plus REST API and Datasette browser for humans.
source_document
Source-linked
Country-scoped from day one. Starting with CH and LI.
bpg · lei
Stable identifiers
Every bank, product, and source document carries a permanent opaque BPG… code.
bank_membership
Memberships
547 bank-membership rows across 13 industry bodies and market infrastructure providers.
Show the data model diagram
Three ways to query the dataOne database, three surfaces. Pick the one that fits your tool.
MCP · For AI agents
mcp.bankproducts.info
Streamable-HTTP MCP server
9 tools — search products, compare banks, list fees and rates
Public data, new structureBank fee schedules are public documents. The factual data they contain is not copyrightable. What we create is a new, structured compilation of these public facts, licensed under CC-BY-4.0.
Trust
Trusted by design
Every fee links to a source page in the original bank document.
For banks
Publish your fees as /.well-known/bankproducts.json
Place a single JSON file on your website. AI agents, comparison tools, and regulators can discover it automatically.
1'302 Fineprint-Regeln aus 139 Schweizer und Liechtensteiner Banken: PostFinance publiziert Diaspora-Preise, ABS macht ESG günstiger, die Hypothekarbank Lenzburg belohnt Aktionäre mit 475 Basispunkten, Grossbanken rabattieren Vermögen, Neobanken rabattieren Volumen, neon verbietet den Wegzug ins Ausland, Revoluts Wochenende beginnt nach New Yorker Zeit, Raiffeisen indexiert eine Preisschwelle an die AHV-Rente. Elf Muster, die zeigen, wie unterschiedlich Banken denken.
Over four weeks we extracted nearly 1'000 footnotes from 60 Swiss and Liechtenstein bank price schedules. PostFinance publishes special prices for the Balkan diaspora, ABS and Glarus Regional Bank make ESG cheaper than conventional, Hypothekarbank Lenzburg rewards shareholders with 475 basis points interest premium, VP Bank offers the same advisory package in three pricing models. Five patterns that reveal how differently banks actually think.
Three real near-misses from an LLM-populated database project — an invented company name, 11 fabricated commercial-registry numbers, 30 banks with their parent cooperative's LEI. The common root is one rule that has to be normative in any LLM-populated data spec, or LLMs will silently ignore it.
In late March 2026 we proposed /.well-known/bankproducts.json as a standard — one JSON file per bank at a known URL. After 4 months and 420 loaded banks we show what the schema actually needs, why zero banks published one, and why the most important rule is „null > invented value".
240 products at 64 Swiss banks had no published price or interest rate in our database. What became visible about Swiss bank transparency — and why we didn't make up a number. Data check from bankproducts.info.
Your products. Your website. One line of code.Embed an interactive product explorer on your bank's website or e-banking portal.
Community tier is live now — Datasette, MCP and REST API all up and free.
Interested? We'll generate a free demo widget with your bank's data.Request a demo
Bankmaster plus — the commercial tierAn expanded dataset for institutional buyers — compliance teams, treasury desks, fintech infrastructure, sovereign analysts. In development; design partners welcome.
Additional attributes under consideration
Direct contacts
Crypto custody addresses
Credit ratings + outlook
Financial scale
ESG positioning
Historical fee + rate snapshots
Change webhooks
Enriched payment-rail data
Supervisor + governance
Interested in being a design partner or getting early access? Pricing is being shaped around real use cases — your input shapes the roadmap.