Rapid self-improvement for your hardware algorithms.

Use large-scale AI agent teams to rapidly discover code improvements on your hardware systems.

$curl -fsSL https://onyxresearch.ai/install.sh | bash

Agent teams that research with you.

With one command, point Onyx at your codebase and have full control over a parallel agent team to rapidly improve any metric.

1
Install Onyx for your Claude, Codex, or OpenCode.
One-line install for the Onyx agent and CLI:
$curl -fsSL https://onyxresearch.ai/install.sh | bash
2
Run /onyx inside your codebase.

Experiments as code.

Onyx uses Git in your existing code to experiment, so your compounding results are seamlessly tracked and ready to be pulled in.

Run Onyx agents anywhere.

Use your same tools

Keep your existing repositories, simulators, and hardware testing workflows. Simply point Onyx at your problems.

Git experiment tracking

Onyx experiments are simply Git code commits with stored metric data, making your development seamless.

Experiment	Name	Metric	Status
a3f92c	Stiffer rear damping	0.913	Succeeded
7c48be	Longitudinal slip term	0.897	Succeeded
e91d07	Thermal derating map	0.874	Succeeded
52b6f9	Anti-windup clamp	0.869	Succeeded
c07a31	Regen torque blend	0.841	Succeeded
91f4d2	Lookahead gain sweep	0.828	Succeeded

Shared team knowledge graph

Every experiment across your whole team becomes durable, compounding knowledge. Breakthroughs transfer across your hardware systems and teammates share learnings.