Agents OS
An AI-powered desktop workspace that keeps chat, tasks, terminal, tools, git context, and local agent execution in one workflow.
Highlights
- Designed the product around a bring-your-own-key workflow.
- Combined chat, tasks, terminal, tools, and git context into one workspace.
- Focused the experience on local execution and low-friction agent workflows.
Context
Agents OS started from a practical problem: AI work often happens across too many surfaces. A user might chat in one place, track tasks somewhere else, run commands in a terminal, and manually move context between them.
The product explores a tighter workspace where AI assistance, execution, and local project context can live together without forcing users into a hosted-only workflow.
What I focused on
I worked on shaping the core product direction around local-first control, bring-your-own-key access, and a workspace model that keeps chat, task state, terminal execution, and tool calls close to each other.
The important constraint was trust. The interface had to make powerful AI actions feel legible, recoverable, and grounded in the user's own environment.
Outcome
The result is a desktop workspace for people who want AI to help with real work while keeping control over keys, context, git state, and local execution.