The framework for agentic analytics

Onyx is an open-source framework for agentic analytics. It is declarative by design and written in Rust. Onyx is built with the following product principles in mind: open-source, performant, code-native, declarative, composable, and secure.

Agentic analytics applies software development lifecycle principles to AI-driven data analytics. Just as traditional software follows a build-test-deploy pipeline, agentic analytics establishes a structured workflow for data agents, involving agent creation, prompt testing, and production deployment.

To learn more, read our docs.

Quickstart

To install Onyx from binary, run the following command (Mac, Linux, WSL):

bash <(curl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/onyx-hq/onyx-public-releases/refs/heads/main/install_onyx.sh)

And to initialize a repo, run the following in the directory you want to instantiate as an onyx repository:

onyx init

At this point, you should have a rudimentary onyx instance set up. To test this out, you can run the following from the root of the directory:

onyx run sql-generator.agent.yml "what is the capital of France?"

See our docs on how to modify the agent file, seed it with context, run tests, and create workflows.