Docs

Model humans.

These docs cover the API: a tokenizer that turns items, actions, or contexts into discrete codes, models that predict and embed over those codes, and an evaluation suite that measures whether any of it holds up against what people actually did. Jean also supplies the behavioral data underneath: sourced, collected, structured to one schema, and delivered with a fidelity report and rights written for foundation models. That part is arranged directly with the team. Talk to us about data.

Tokenize anything

Fit a tokenizer on items, behaviors, or contexts. Get short integer codes back.

Models

Predict the next item and embed a history, on an assigned deployment.

Evals

The public board, and how to get your own system run against it.

API reference

Authentication, errors, jobs, and every endpoint.

Access is provisioned with the team. Onboarding supplies a base URL, an API key, and the capabilities enabled for your workspace. These docs define the contract; examples illustrate its shape. Request access.

Make your first call

Start by giving the tokenizer something to encode. Anything with text works: products, articles, actions, situations, survey items. Jean returns a short sequence of integer codes per entry, and similar entries share a prefix.

curl --fail-with-body -X POST "$JEAN_BASE_URL/tokenizers" \
  -H "Authorization: Bearer $JEAN_API_KEY" \
  -H "Idempotency-Key: YOUR_UNIQUE_REQUEST_KEY" \
  -H "Content-Type: application/json" \
  -d '{"dataset_id":"$JEAN_DATASET_ID","levels":8}'

The response is a job. Poll it, then read the codes. The whole flow, including how to get a dataset in, is on Tokenize anything.

Start

Build

Evaluation

What Jean does

Model humans.

Tokenize anything

Models

Evals

API reference

Make your first call

​Model humans.

Tokenize anything

Models

Evals

API reference

​Make your first call

Model humans.

Make your first call