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/refs/heads/main/install_onyx.sh)

Alternative Installation Methods

Using Homebrew (macOS only)

brew install onyx-hq/onyx/onyx

Installing a Specific Version

ONYX_VERSION="0.1.0" bash <(curl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/onyx-hq/onyx/refs/heads/main/install_onyx.sh)

To verify the installation, run:

onyx --version

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 commands from the root of the directory:

onyx run sql-generator.agent.yml "On how many nights did I get good sleep in the last year?"  # ask a question to the sample agent
onyx test sql-generator.agent.yml  # runs all defined tests against the sql-generator agent

onyx run report-generator.workflow.yml  # execute the sample workflow
onyx test report-generator.workflow.yml  # run all defined tests against the workflow

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