Agentic EDASource on GitHub
Credit-risk EDA with verified AI reports

Exploratory analysis you can trust

Upload a CSV and get a report on what drives your outcome. Statistics find the insights, the AI only writes them up, and a checker verifies every number against the data.

Free · no dataset needed to try it
ExampleFrom a report on the synthetic credit dataset, not live data
sample_credit_applications.csv Every number verified

Credit score is the strongest applicant signal: 26.2% default in the bottom 20% vs 3.0% in the top 20%.

low credit scorehigh credit score
IV
0.63
KS
0.32
Gini
0.43
100%
planted risk drivers found
held-out data, 5,000+ rows
2.4%
false-discovery share
target ≤ 5%
Every
number checked
by a deterministic critic
400+
automated tests
real Postgres + Redis in CI

Why it's different

AI that can't make up numbers

Statistics find the insights

Every column is tested with false-discovery control and an effect-size floor, so noise never becomes a finding.

Every number is verified

The AI only writes up verified findings. A checker rejects any number that isn't in the evidence for that variable.

Built for credit risk

Information Value, KS curves, lender-decision variables and fair-lending proxy checks, out of the box.

How it works

From CSV to verified report in seconds

  1. STEP 1

    Upload a CSV

    Or try a synthetic credit dataset. Your file is deleted as soon as it's read.

  2. STEP 2

    Watch it analyse

    Live progress: planning, statistics, writing, checking.

  3. STEP 3

    Read a verified report

    Findings, interactive charts, scorecard metrics and fairness warnings.

See it on a real credit dataset

Sign in and press “Try a sample credit dataset”: a full report in under a minute, nothing to upload.