I built Vificil to use browser AI with local codebases — without an API

Özgür Daylan — Mon, 31 Aug 2026 13:55:13 +0000

I use browser-based AI tools like ChatGPT and Claude a lot while developing software.

The models are capable, but I kept running into the same annoying workflow:

Collect the relevant project files.
Upload them to the AI.
Explain what needs to change.
Copy generated code back.
Figure out which files changed.
Apply everything manually.
Run the project and hope nothing was missed.

For small changes this is manageable.

For multi-file changes, it gets old very quickly.

So I built Vificil.

The idea

Vificil connects browser-based AI tools with your local project without requiring the AI to have direct access to your filesystem.

Instead of copying individual files and code snippets back and forth, Vificil prepares a cleaned Full Source Context of the project.

You give that context to the AI you're already using and describe the change you want.

The AI returns a structured patch ZIP.

You download it.

Vificil detects it and handles the rest.

It can validate the patch, apply the changes to the correct project, run verification commands, keep backups, and roll back failed changes.

So the workflow becomes roughly:

Local project → Vificil → Browser AI → Patch ZIP → Vificil → Local project

Why browser AI?

A lot of AI coding tools are built around APIs, integrated agents, their own IDEs, or local models.

Those are useful approaches, but I wanted something different.

I wanted to keep:

my existing IDE
my local project
the AI service I already use
my existing consumer AI subscription

Vificil itself doesn't require an API key or local-model hardware.

For example, someone already using ChatGPT on the web can use that existing access for the reasoning and code generation part of the workflow instead of necessarily routing every task through separately billed API calls.

It's not tied to ChatGPT or Claude

ChatGPT and Claude are the obvious examples, but Vificil is designed to be AI-independent.

The AI doesn't need a special Vificil integration.

If it can understand the supplied project context and generate the structured patch ZIP Vificil expects, it can potentially be used with the workflow.

That also means Vificil isn't trying to become another AI IDE.

The goal is closer to:

Use your AI. Use your IDE. Keep your project local. Let Vificil handle the handoff between them.

No more manual patch copying

One part I particularly wanted to eliminate was this:

AI:

Replace this function with...

Then manually opening the file, finding the right location, copying the code, checking imports, moving to another file, repeating...

With Vificil, the AI describes those changes through the patch format instead.

The downloaded ZIP becomes the handoff.

Vificil can then apply it locally and verify what happened.

Download: https://apps.microsoft.com/detail/9P0JC8VVQB75

Watch the tutorial

Here’s a short walkthrough showing how Vificil prepares project context, receives a patch ZIP from the AI, and applies it back to the local project:

**Free version

Vificil Basic is free with no usage quotas.**

I'm launching it now because I want to find out whether other developers have the same problem I had.

I'm especially interested in feedback around:

the browser-AI → local-project workflow
Full Source Context
the structured patch approach
verification and rollback
which AI tools people would like to use with it

Launch offer 🎁

I'm also giving Vificil Pro free to the first 100 users through a Microsoft Store promotional link.

http://go.microsoft.com/fwlink/?LinkId=532540&mstoken=WFRJ3-QTJH6-KXJCV-4WXG9-HDM3Z

No promo code is required. The offer is limited to the first 100 redemptions.

If you try Vificil, I'd genuinely like to know where the workflow feels good — and where it still gets in your way.

DEV Community: Özgür Daylan

I built Vificil to use browser AI with local codebases — without an API