flagship · 1.0 · project
Omnitron
AI coding agents in your own workspace
My build of DeepSeek Harness, the open-source platform for running AI coding agents in your own workspace: one plugin core behind a web UI and a desktop app for Linux, macOS and Windows, released as Omnitron 1.0 with its own identity, interface and operational guards.
The problem
Coding agents mostly arrive as someone else's product running in someone else's cloud. DeepSeek Harness, an MIT-licensed platform, runs them in your own workspace instead, on hardware and files you choose. Omnitron is my build of it.
How it works
- Everything is a plugin: the agent runtime, its tools and its model providers compose on the Cordis framework, so swapping a provider or adding a tool is a plugin rather than a fork.
- One web UI, two ways in: a single CLI command serves it locally, and a desktop app packages the same UI and runtime for Linux (x64 and arm64), macOS (arm64) and Windows (x64), with no Node.js to install.
- State stays on the machine: the desktop app and the CLI keep their data in the user's home directory, not on a service.
- My 1.0 on top of the upstream project: a new identity and interface, and operational guards for running it day to day. It is still experimental developer-preview software, and says so.
Why it exists
Where an agent runs decides who sees the code. It is the same argument as Wayne and PrometheusAI: capable AI on hardware you control, with the controls in view.
More work
Hiring for AI or ML?
I am open to AI/ML Engineering, Data Science, and Python roles, plus research collaborations and consulting. New York based, shipping worldwide.