Hariom AI

A Windows-first personal AI workstation designed to plan tasks, use tools under controlled permissions, and verify what it does.

View source

01 / THE PROBLEM

Why this exists

Desktop assistants can feel powerful while making it difficult to understand what they are about to change. Hariom AI is being built around explicit tool boundaries, reviewable actions, and verification.

02 / PRODUCT DIRECTION

Where it is going

A local-first assistant experience with structured plans, approval gates for sensitive actions, bounded execution, and a clear record of outcomes.

03 / CURRENT SCOPE

What we can say today

Structured tool registry and task plans
Approval-gated sensitive operations
Browser and screen-control experiments
Test and verification workflows

This page describes the current project intent. It does not imply that every planned capability is released or production-ready.

BUILD SOMETHING TOGETHER

Have a related problem?

Let’s discuss the goal, constraints, and a sensible first version.

Discuss a project