Credit Readiness Advisory
Diagnostic engine for German and Austrian SME credit files.
It answers one question: if this file went to a lender today, which factors
would pull it below an approvable line — and which of those are fixable before
the client ever reapplies?
This is not a credit rating tool. It produces a directional readiness
indication, never a rating or a probability of default. See
docs/regulatory-guardrails.md — that boundary
is enforced in code and covered by tests.
Quick start
See it working (demo): double-click start_demo.bat, or run
python -m credit_readiness serve --demo --open # http://127.0.0.1:8766 via the .bat
This starts three pages with four fictional companies:
| Page | For whom |
|---|
/ | the public website: what we do, how it works, pricing (DE/EN) |
/investors | investor one-pager: problem, solution, market, model, status |
/app | the portal: log in or register, then a view per role |
Demo logins (password demo1234, also shown as one-click buttons on the login page):
Delete data/demo/ to reset the demo; it is re-created on the next start.
Real use: double-click start_portal.bat (data in data/clients/). Create your
advisor account first:
python -m credit_readiness user add [email protected] "Your Name" --role berater --password "..."
Companies then register themselves on the website, or you create a case and invite them.
The portal is the everyday tool: create a case, fill in or import the two
questionnaires, upload documents, generate the request letters, run the
diagnostic, print the report, record the outcome. Everything it does is also
available from the command line:
python -m credit_readiness case new "Firma GmbH"
python -m credit_readiness case answers CRA-2026-0001 unternehmen antworten.json
python -m credit_readiness case upload CRA-2026-0001 susa_aktuell susa.csv --period-end 2025-12-31
python -m credit_readiness case upload CRA-2026-0001 kontoumsaetze konto.csv
python -m credit_readiness case letters CRA-2026-0001
python -m credit_readiness case diagnose CRA-2026-0001
python -m credit_readiness case outcome CRA-2026-0001 APPROVED_BETTER_TERMS --amount 750000
python -m credit_readiness questionnaire unternehmen --format html -o fragebogen.html
python -m credit_readiness documents # document checklist
python -m credit_readiness bank konto.csv --limit 500000 # analyse a bank export
To see a complete engagement run end to end on fictional data:
python scripts/generate_intake_samples.py # questionnaires, DATEV exports, bank CSV, PDFs
python scripts/demo_intake_flow.py # case -> uploads -> letters -> diagnostic -> outcome
Hand-built cases and the batch mode still work:
python scripts/generate_samples.py
python -m credit_readiness batch data/samples
python -m credit_readiness diagnose data/samples/case_01_mueller_praezisionstechnik.json -o output/
Sample output:
Fall Unternehmen Band Score -> Score Band Einordnung
CASE-01 Mueller Praezisionstechnik B 66.4 -> 80.3 A behebbar - Struktur
CASE-02 Nordlicht Handel GmbH & Co. D 50.7 -> 64.7 C behebbar - Struktur
CASE-03 Gastro Rheinblick GmbH E 4.2 -> 8.7 E substanzielles Kreditrisiko
CASE-06 Hoffmann Medizintechnik A 93.3 -> 93.3 A bereits finanzierbar
Products
| Tier | Price (indicative) | What happens |
|---|
| Credit check | free | Company uploads its annual accounts, the reader (local rules or Claude) proposes the figures, the company confirms them, answers ~12 questions, and gets an instant readiness band + top 3 weaknesses with a plain-language explanation |
| Full report | EUR 390 | Complete questionnaire and documents, full diagnostic, reviewed and released by an advisor |
| Advisory | from EUR 1,500 | Calls, Steuerberater sign-off, support up to the bank meeting |
AI reads and explains; transparent rules decide. See
docs/ai-and-data-protection.md. By default the
local reader is used and no data leaves the machine. To use Claude:
pip install "credit-readiness[ai]" # anthropic SDK
set CRA_AI_PROVIDER=anthropic # plus Anthropic credentials (ANTHROPIC_API_KEY)
Only do this after the data processing agreement with the provider is in place.
What it does
Intake (see docs/inputs-outputs.md for every input and output):
This is the beginning of the README. Read the rest at the original. Read the full README ↗
