For years, writing a data quality check meant opening the docs and figuring it out the syntax yourself.

These days software enginering is changing, we are more confortable at letting agents drive while we sit as passengers. But data quality is still written by hand, the checks, the contracts, the data source configs. The terminal has gotten smarter everywhere else. But not here.

Today we're changing that.

Soda AI CLI is a data quality agent for your terminal. It knows Soda's syntax, data quality best practices, it reads your actual files, and takes action on them. Write contracts from scratch. Migrate from checks to contracts. Create full data contracts from pipeline or dbt scripts. All without leaving your terminal.

It ships today as part of soda-core. Open Beta. No extra install, no API keys.

Just run soda ai.

Introducing Soda AI CLI

soda ai is a data quality agent for your terminal.

Not a chatbot. An agent. One that knows data quality deeply, reads your actual files, and takes action on them. No switching contexts, no browser tabs, no copying YAML into a chat window.

It ships today as part of soda-core, our open-source data quality engine. No extra installation, no API keys. Just upgrade and run:

Using UV (recommended)

#replace soda-{data source} with your data source.
uv pip install git+https://github.com/sodadata/soda-core.git@soda-ai#subdirectory=soda-postgres

Using pip

#replace soda-{data source} with your data source.
pip install git+https://github.com/sodadata/soda-core.git@soda-ai#subdirectory=soda-postgres git+https://github.com/sodadata/soda-core.git@soda-ai#subdirectory=soda-core

What Soda AI CLI can do

Write contracts from scratch

Give soda ai your table structure, a dbt model, or a pipeline script, and it will draft a complete data contract: schema definitions, quality checks, freshness rules, thresholds. Not a template to fill in. A contract you can deploy.

Translate ODCS contracts to Soda format

If you're coming from ODCS and want to use Soda's enforcement engine, soda ai translates your existing contracts into Soda Contract Language automatically. You review, you approve, it saves the file.

Migrate v3 check files to v4 data contracts

The v3-to-v4 migration is a different syntax, a different mental model, and a list of edge cases you'd rather not discover manually. soda ai handles the translation. Reference a file by name in the session and it reads it directly.

Answer data quality questions in context

Not generic advice. Answers that account for what you're actually trying to do, in Soda's syntax, against your actual setup. "What's the right check for detecting when a percentage column goes above 100?" or "Why is my freshness check failing even though the table updated?" are the kinds of questions it's built for.

What a session looks like

You reference files by name. Soda AI reads them. Translated contracts get saved directly to disk, with your confirmation.

Why terminal-native matters

The terminal is where data engineers live. It's where pipelines run, where configs live, where debugging happens. Building soda ai as a CLI agent wasn't just a design choice. It was a statement about where data quality work actually happens.

Not in a browser tab you open once a month. In the same environment where you already work.

Talk to your data quality layer the way you talk to your codebase.

This is experimental. And that's intentional.

We're shipping soda ai under an experimental flag because we mean it. The agent will improve significantly based on how real data engineers actually use it. We're not waiting for it to be perfect to share it. We want your failures as much as your wins.

Data contracts are what it specializes in today. That's not the ceiling.

The direction we're building toward: a data quality agent that understands your full stack. Your orchestration, your pipeline definitions, your existing quality rules, your failure patterns. One that proactively surfaces what needs attention and helps you act on it, without leaving your terminal.

We're building it in public. And we want to hear from you.

Try it today

#replace soda-{data source} with your data source.
uv pip install git+https://github.com/sodadata/soda-core.git@soda-ai#subdirectory=soda-postgres

Check the documentation for the full setup guide. Soda AI CLI is part of Soda Core, our open-source data quality engine. It is free to use under our fair use policy.