Skip to content
Onyx

Agent-in-the-loop hardware testing.

Let your coding agents evaluate ideas with hardware-in-the-loop testing
to discover engineering improvements at ridiculous speed.

How it works.

Onyx hardens your hardware-in-the-loop (HIL) and software-in-the-loop (SIL) testing setups for coding agents and rapid self-improvement.

  1. Step 01
    Step 01

    Agent-ready HIL and SIL

    Use your own coding agents with hardened guardrails, remote job orchestration, and sandboxing at scale.

  2. Step 02
    Step 02

    Scalable knowledge graph

    Edge relations connect across your team so agents leverage all data without running redundant experiments.

  3. Step 03
    Step 03

    Self-improvement algorithm

    We combine AI creativity with mathematical optimization to propose new experiments and self-improve.

A new engineering capability.

Specify goals with complete control over your own coding agents as they run massive lights-out HIL and SIL experiments to solve critical problems.

Onyx's Team
GoalUAV battery dynamics model accuracy
Onyx's Team
GoalUAV battery dynamics model accuracy
1 agents0 experiments0 hypothesesbest —

Built for any hardware.

Use your own custom hardware, software tools, and coding agent interfaces.

GoalLearn the motor's dynamics from bench data

Security

TYPE IITYPE II

Security

SOC 2 Type II compliant

Open source

Open source

Open source, built by engineers

Compatibility

Compatibility

Keep your engineering tools

Use anything your coding agent can.

Be ambitious with your engineering.

No tutorial, just ask your coding agent to set Onyx up in minutes.

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