An independent software initiative

Thoughtful tools.
Real work.

We build practical software for AI-assisted work and engineering workflows. Local-first by design. Built around human control.

Early-stage projects. Clear scope. Honest progress.
// 01Local-first thinkingKeep ownership close to the user.
// 02Human controlMake consequential actions explicit.
// 03Evidence over promisesSeparate what works from what is next.
01 / Selected work

Small tools. Useful foundations.

From controlled agent automation to clearer assembly instructions. Each project solves a specific workflow problem.

agent-safe-runner

Published · Alpha

A local Python job runner for AI-proposed commands. Explicit approval, policy-controlled execution, a durable queue and an auditable event history.

Explore on PyPI
Open source · MIT license · An execution gate, not an OS sandbox

AssembleLive

Local prototype

Turn instruction images into editable, step-by-step assembly guides. A bilingual foundation with Vietnamese and English captions, local saving and a versioned project format.

Talk about the project
In development · AI analysis and 3D playback are planned

MessengerBubbles

Released · v0.6.1

Floating Facebook chat bubbles for Windows 11, one bubble per conversation, like chat heads on Android. A WPF and WebView2 desktop app with message previews, sounds and a Vietnamese or English interface.

View on GitHub
Windows 11 · Personal project, not affiliated with Meta
02 / On the workbench

Questions worth building for.

These are discovery-stage concepts, not launched products. We are shaping their scope around real workflows.

01 / PLANNING

ShiftFair

Excel-compatible shift planning for small teams. Preserve locked shifts, reduce disruptive changes and make trade-offs understandable.

Discovery
02 / DIAGNOSTICS

NetLens

Network troubleshooting guided by bounded measurements, explicit uncertainty and a clear next test. Evidence before automatic fixes.

Discovery
03 / HANDOFFS

Project Passport

Creative-project handoffs with evidence of what was packaged and tested. Starting with Blender dependencies and environment checks.

Discovery
03 / Our approach

An engineering mindset.
A practical purpose.

Founded and developed by Nguyen Xuan Thanh, a mechanical engineer based in Hanoi, Vietnam. For enquiries and pilot discussions: xuanthanh@umefor.space.

Umefor is an early-stage, independent software initiative shaped by a mechanical engineer’s approach to problem-solving: understand the task, define the constraints, build a small working system, and test it.

We explore how AI can help people interpret information and move through complex work, while keeping validation, data ownership and important decisions in human hands.

  • 01
    Start with a specific workflow, not a universal platform.
  • 02
    Keep model suggestions separate from verified results.
  • 03
    Make progress inspectable and limitations visible.
04 / Say hello

Have a workflow worth improving?

For project enquiries, feedback or a conversation about a potential pilot, get in touch directly.

xuanthanh@umefor.space