ChatGPT To Codex
A local MCP runtime for macOS and Windows that enables ChatGPT to read, edit, search, run tests, and capture screenshots in your local project folder while keeping your source code on your machine.
README
<p align="center"> <img src="assets/readme-hero.png" alt="ChatGPT To Codex local coding runtime" width="100%" /> </p>
ChatGPT To Codex
Give ChatGPT real local coding hands.
ChatGPT To Codex is a local MCP and Actions runtime for macOS and Windows that lets ChatGPT work inside the project folder you choose: read files, search code, apply patches, run tests, launch E2E checks, and send back screenshot proof.
Your source stays on your machine. ChatGPT connects to the local app you run. You choose the workspace, approve the token, and keep control of what gets edited.
Download v0.1.2 · Beginner installation guide
Help us get this in front of more builders: star the repo if you want ChatGPT to stop talking about code and start safely doing the repo loop.
Why It Exists
ChatGPT is great at reasoning, but web chat alone cannot reliably inspect your local repo, run your local tests, or prove what the UI actually looked like. ChatGPT To Codex fills that gap:
- local project selection instead of uploading a source tree
- guarded file reads and hash-checked patching
- allowlisted local commands for tests and checks
- macOS/Windows app, window, and browser screenshot capture for visual E2E proof
- temporary or fixed HTTPS connector URL for ChatGPT web
- OAuth-style owner-token approval so random clients cannot just attach
- multilingual menu bar app for non-English users
The mental model is simple:
ChatGPT thinks. Your computer acts. You review the result.
Current Release
| Platform | Status | Package |
|---|---|---|
| macOS | Public release | chatgpt2codex-0.1.2.pkg |
| Windows | Public release | chatgpt2codex-0.1.2-windows-setup.exe |
| Linux | Developer path only | Not published |
Why PKG Instead Of DMG?
For this release, PKG is the better fit. A DMG is nicer for drag-and-drop apps, but this app needs to install a menu bar runtime under Applications, bundle Node/cloudflared helpers, and run a non-blocking post-install Doctor. PKG gives beginners a clearer "install and open" path. A signed/notarized DMG can still be added later for a more consumer-style download.
Current macOS package SHA-256:
d9226ba10d9a24a121a0ec6c7a6c79f81d55b3bb60ddc5acd13240c2d17f7821 chatgpt2codex-0.1.2.pkg
What ChatGPT Can Do With It
Once connected, ChatGPT can operate like a practical coding agent over a trusted project:
- list local projects and select the active one
- read repo rules before editing
- search code and read exact line slices
- create files and apply patches
- run project commands and tests
- start a dev server and wait for a URL
- open a browser URL or installed desktop app
- capture macOS/Windows E2E screenshots
- return inline screenshot previews through Actions
- save generated image assets into the repo
- summarize diffs, blockers, and verification evidence
The standout workflow is:
Run the E2E test, open the app, capture screenshots, and show me proof.
For web apps, ChatGPT To Codex can capture browser regions. For desktop apps
such as Tauri apps, it can open the built app window and capture top/middle/bottom
views. The one-shot e2e_test_and_show_screenshot action returns inline
imageMarkdown results so you can inspect the screen without digging through
local folders.
Install In 5 Minutes
Full beginner guide: docs/INSTALL.md
macOS short version:
- Download
chatgpt2codex-0.1.2.pkgfrom the latest release. - Open the installer.
- If macOS blocks the unsigned package, Control-click it, choose Open, and confirm in System Settings -> Privacy & Security if needed.
- Open ChatGPT To Codex from Applications.
- Open Settings... from the menu bar icon.
- Choose a project folder.
- Enable ChatGPT web connector if you want ChatGPT in the browser to connect.
- Click Start MCP.
- Click Copy Connector URL.
- Register that
/mcpURL in ChatGPT Apps / Connectors and approve with the Owner Token shown by the app.
Windows short version:
- Download
chatgpt2codex-0.1.2-windows-setup.exefrom the latest release. - Double-click the installer.
- If Windows SmartScreen warns, choose More info -> Run anyway only if the file came from this GitHub release.
- Launch ChatGPT To Codex.
- Open the tray icon settings, choose your project folder, enable the ChatGPT web connector if needed, then click Start MCP.
- Copy the
/mcpConnector URL and approve it in ChatGPT with the Owner Token.
Keep the Owner Token private. Treat it like a password.
First Prompt To Try
Use ChatGPT To Codex. Select my project, read the README and package scripts,
run the safest available check, then summarize the result with exact evidence.
Then try a visual proof flow:
Use ChatGPT To Codex to run the app E2E, capture screenshots, and show the
passing screenshot set inline before you say it is done.
Safety Model
ChatGPT To Codex is designed for trusted local development, not arbitrary public automation.
- It runs locally on your computer.
- It defaults to loopback-only networking.
- ChatGPT web requires an explicit connector/tunnel mode.
- File operations are scoped to the selected project.
- Patch application uses line/hash context.
- Owner Token approval is required for remote Actions access.
- Secret-looking values are redacted from tool output.
- Destructive, network, and sensitive operations remain approval-gated.
Do not expose the connector URL publicly unless you understand the tunnel and token model. Do not paste Owner Tokens into issues, screenshots, or shared logs.
Supported Languages
The desktop app can follow the system language and currently includes UI strings for English, Korean, Japanese, Simplified Chinese, Traditional Chinese, Spanish, French, German, Brazilian Portuguese, Italian, Dutch, Polish, Russian, Turkish, Vietnamese, Indonesian, Thai, Arabic, Hindi, and Ukrainian.
The install guide currently includes Korean, English, Japanese, and Simplified Chinese. More documentation languages are welcome.
Windows Status
Windows now has a public beginner installer. It includes the tray launcher, owner-token setup flow, ChatGPT web connector settings, stale runtime cleanup, and Windows E2E screenshot proof. See docs/INSTALL.md and windows/README.md for the full Windows guide.
Repository Contents
This public repository is intended to contain only the product source, public documentation, assets, scripts, and published installer artifacts. Local agent state, personal automation rules, generated memory, hooks, private MCP config, build output, and machine-local logs are ignored.
If you see local-only files in a clone, they came from your machine, not from the public repo.
Build From Source
For developers:
npm ci
npm run typecheck
npm test
npm run build
Build the macOS package:
npm run macos:package
The packaging script creates a .pkg under build/macos/. Published packages
are copied to installers/macos/.
Star Pitch
If this saves you one "copy this patch, paste it in terminal, now run tests, now send me a screenshot" loop, give it a star. The goal is simple: make ChatGPT useful for real local development without turning your project into a cloud upload.
Built by ezBuilder.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.