NEW

Dex for AI analytics engineering and dbt

Give your coding agent a repeatable way to explore data, build dbt models and maintain metrics.

Updated November 17, 2026

Screenshot of Dex

Key facts

What it is
Give your coding agent a repeatable way to explore data, build dbt models and maintain metrics.
Pricing
Free
Category
AI Coding & Developer Tools
Website
github.com

Dex is Exmergo’s open-source analytics engineering toolkit for AI coding agents. It gives an agent a structured way to understand a warehouse, propose dbt and semantic-layer changes, and check what breaks when source data changes. Its outputs include profiles, JSON results and reviewable project files.

What does Dex actually do?

Imagine asking an agent to build a revenue model. Writing SQL is only part of the job: it must identify the right tables, check whether customer keys match, understand what one row represents and avoid double-counting. Dex supplies that exploration and verification workflow around the agent.

The repository is useful to analytics engineers and data engineers who work with SQL, dbt and Git. You run its skills or CLI locally; it requires a shell-capable agent and uv. A normal browser chat is not the execution environment.

Inspect tables and verify joins before building

Explore maps relevant objects, profiles columns and reports candidate keys, personal-data flags and data-quality findings. Verify inferred relationships before relying on a shared column name. An ER diagram helps explain the structure, while the measurements help decide whether a join is trustworthy.

Dex keeps source exploration read-only. An inferred key or relationship is evidence to examine, rather than a guarantee that the business meaning is correct.

Try a complete local DuckDB workflow

Start in a fresh directory outside an existing Dex project, with uv available. Follow the official quickstart:

uv tool install "exmergo-dex-core[duckdb]"
mkdir dex-first-run
cd dex-first-run
dex init analytics --mode demo
dex explore map
dex explore profile order_items products
dex explore relationships --verify
dex explore query "select count(*) as order_count from orders"

Installation downloads packages; the demo then runs locally without your own warehouse credentials. Initialization creates a DuckDB file, repository configuration and a dbt skeleton. Read the JSON status, data, warnings and error fields instead of assuming initialization means a successful build.

The demo deliberately includes duplicate order-item keys and a customer relationship that fails verification. Those findings are expected. The final aggregate query returns 5,000 orders, giving you a concrete result to check.

Review dbt models, semantics and drift repairs

Transform prepares changes to models, tests, documentation and metrics as stored plans. Review the diff, apply the intended change and build on a development target. Dex Semantics records grain, population and aggregation assumptions alongside a model, then checks built data against those declarations.

Maintain compares the project with a snapshot to find schema, volume, grain and metric-definition drift. It proposes repairs for review. Without a baseline, maintain verify can still inspect the project as it stands.

Check database and semantic-layer compatibility

Connectors include DuckDB, PostgreSQL, Snowflake, BigQuery, Databricks, Redshift, ClickHouse and Trino. Access to MySQL, SQL Server or MongoDB is documented through Trino catalogs; do not assume native connectors.

Semantic backends have different capabilities. dbt, Cube, SLayer and Apache Ossie do not all support the same queries. Authoring a definition is not equivalent to deploying or querying it. Developers can also embed the engine through the DexEngine Python API.

Free software, outside costs and privacy boundaries

Checked November 16, 2026: open-source Dex is free under Apache 2.0, with no account or credit card required. Model-provider and warehouse charges remain separate. Official pricing describes Enterprise as a separate custom-priced offering in development with design partners.

Metered work uses estimates, confirmation and configured ceilings. Estimates are not billing guarantees, and the hosted dbt Cloud Semantic Layer has no Dex-enforced ceiling; consult the cost-control reference.

The open-source privacy policy says Dex sends nothing to Exmergo. Your agent may still send returned context to its model provider. PII detection is heuristic; caches and snapshots can contain bounded nonpersonal values, and SQL logs can contain filter literals. Review .dex/ before committing or sharing it.

Frequently asked questions

Is Dex free to use?

Yes, the open-source software is free. Your agent, model and warehouse can have their own charges; Enterprise has separate terms.

Can I start without a dbt project?

Yes. Read-only exploration does not require dbt, and the demo creates a starter project and local warehouse.

Does Exmergo receive my warehouse data?

The open-source policy says no. That does not mean your model provider receives no context or that local files contain no business values.

Official resources and developer community

Alternatives to Dex

Visit github.com
Pricing
Free
Website
github.com
Added
November 2026
Updated
November 17, 2026

More in AI Coding & Developer Tools

Sponsored

Promote your AI tool here

Reach people comparing tools in this category.

See options
Is this your tool?Update the listing
Add the badgeShow you're listed on AITroveX
Listed on AITroveX
<a href="https://aitrovex.com/tool/dex?ref=badge" target="_blank">
  <img src="https://aitrovex.com/badge-listed-light.svg" alt="Listed on AITroveX" style="height: 54px; width: auto;" />
</a>