Welt is a collection of the strongest AI models built for tables, from labs including Inria, Amazon, Kumo and Mistral AI. Give one your spreadsheet and ask a question: who's likely to leave, what sales will look like next month, what's driving a result. You get an answer in minutes, with nothing to train.

Which models are in the Welt collection?

TabICLv2, Kumo Tabular Medium, Mitra-v2 and TabDPT 1.3 predict a category or a number. Chronos-2 and Timestral forecast what comes next. Arrow, CDFM and AVICI find cause and effect. More models, including Welt's own, are joining next.

Do I need to train or tune anything?

No. These models have already learned from millions of tables, so they read your rows as examples and predict straight away. There's no training run and nothing to tune.

Can I use Welt from Claude or ChatGPT?

Yes. Add https://welt.ergodic.dev/mcp to Claude, ChatGPT, Cursor or Codex as a custom connector, sign in once in your browser, and ask in plain English. Your assistant runs the models for you.

How do I get started?

With no code, connect your AI assistant. In Python, run pip install "welt-client[parquet]", then welt login. Or open a ready notebook in Google Colab from the Welt docs at https://ergodic-ai.github.io/welt-python/.

Find the signal. Build what’s next.

AI models that answer questions about your tables. Predict, forecast and find what drives your results, in minutes, with nothing to train.

What is Welt?

Welt is a collection of the strongest AI models for tables, run for you behind one API: prediction (TabICLv2, Kumo Tabular Medium, Mitra-v2, TabDPT 1.3), forecasting (Chronos-2, Timestral) and causal discovery (Arrow, CDFM, AVICI). Use it from Python (pip install welt-client), the REST API, the welt CLI, or your AI assistant through the hosted MCP server at https://welt.ergodic.dev/mcp. Try Welt.

ergodic, open source

ergodic is Ergodic’s open-source Python package for causal inference and process intelligence (pip install ergodic, Apache 2.0). It takes the causal graph Welt finds, adds domain knowledge, tests hypotheses and estimates what each action would change, on your machine. Welt itself does not estimate causal effects. ergodic docs.

From a table to a decision, in four steps

Predict: what will happen? In Welt.
Explain: why does it happen? Causal discovery in Welt.
Compare: what would each action change? Effect estimation with ergodic.
Decide: which action to take? Coming soon: welt-jev, the decision model powered by hard data.

The Library: articles and 107 worked use cases on real tables.
llms.txt and llms-full.txt: this site for AI agents, in Markdown.
Welt product docs for agents
Python docs