Our Vision
Every developer deserves an AI teammate that actually works.
One that runs on your machine, respects your privacy, remembers your preferences, and ships code alongside you.
How GrayCode is organized
One agent, a few focused tools
Rho is the terminal AI coding agent and the main GrayCode product. Flux powers its model access, GrayCode Skills extends it, and GrayCode Cloud is an optional hosted control plane.
Agent-neutral tools
Rover (agent workspaces), Across (engineering context and provenance) and Trace (a self-hosted Git forge) are separate open-source tools. They work alongside Rho and do not require it.
Privacy by default
Your code goes to the model provider you choose and nowhere else. Cloud sync is opt-in.
Model agnostic
Anthropic, OpenAI, Gemini, OpenRouter, Ollama and more through Flux. You choose.
Local-first
Rho is a single binary that runs on your machine, in any shell.
Open source
MIT licensed. Every prompt, every tool definition — fully auditable.
Focused tools
Rho is the agent. Flux, GrayCode Skills and GrayCode Cloud power and extend it; Rover, Across and Trace stand on their own.
Built by developers
No investors. No board. We build what we use every day.
The problem
The way we write software is changing faster than the tools can keep up. AI can generate entire files in seconds, debug code it's never seen, and refactor systems across thousands of lines.
But the developer experience around these capabilities is still broken. Juggling tools that don't share context. Sending proprietary code to someone else's cloud. Paying per token for something that should feel as natural as running a linter. Trusting a black box.
“We think that's wrong. And we think it's fixable.”
Our approach
GrayCode is building Rho, an AI coding agent that runs on your machine. Your code goes only to the model provider you pick, with your own key. Your keys stay in your OS keychain. Every line is MIT licensed, because trust is built through transparency, not terms of service.
Rho keeps its tools in one session because AI coding tools need shared context to be useful. When one agent can read the code, run the tests, check the history and review the change, it stops being a toy and starts being a teammate. Around it we build small, focused projects: Flux for model access, GrayCode Skills for reusable workflows, Rover for agent workspaces, Across for engineering context and Trace for self-hosted Git.
Where we're going
Right now we're building Rho in the open. It is alpha: features, tests and hardening come first, and a recommended public install comes when it is ready. The goal is a CLI you can trust every day.
We want to build the platform where solo developers and their agents can bring their own models, their own data, and their own workflows, and have an AI teammate that genuinely understands their codebase. Not a chatbot that guesses. A teammate that remembers, reasons, and ships.
We want AI coding to be as reliable as version control. Something you depend on every day without thinking about it. Something that just works — one bit at a time.
Why “GrayCode”?
The name comes from the binary encoding system where only one bit changes at a time — small, reliable steps forward. That's our promise. We won't move fast and break things. We'll move deliberately and build things that last.
We're two engineers. No investors. No board. No one telling us to ship faster at the expense of quality. We fund this ourselves because we want to build something we're proud of — not something that looks good in a pitch deck.
The future of coding is human and AI, building side by side.
We're here to make that future work.
Read the docs