codex-chatgpt-control  by adamallcock

SDK for agents to control visible ChatGPT web sessions

Created 1 month ago
297 stars

Top 89.2% on SourcePulse

GitHubView on GitHub
Project Summary

<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This project offers an unofficial SDK facade for Codex agents, enabling control over visible ChatGPT web sessions. It addresses the need to integrate Codex's local execution with ChatGPT's advanced reasoning, planning, and research capabilities, solving the manual tab-switching problem. The benefit is a structured, user-directed bridge that leverages ChatGPT's strengths while keeping Codex as the primary environment, without requiring direct API access.

How It Works

The SDK bridges Codex and ChatGPT via a Node.js backend and Python client, driving chatgpt.com through a compatible browser bridge and user-visible UI. It provides workflow primitives for prompts, threads, and response capture, not a generic API wrapper. This approach ensures user control and privacy by operating within a visible session, allowing agents to consult ChatGPT for tasks like deep planning or research synthesis.

Quick Start & Requirements

  • Installation: Node.js (npm install codex-chatgpt-control), Python (python -m pip install codex-chatgpt-control), Codex plugin (codex plugin marketplace add adamallcock/codex-chatgpt-control --ref main followed by codex plugin add codex-chatgpt-control@codex-chatgpt-control).
  • Prerequisites: Node.js 20+, Python 3.10+, npm, a signed-in ChatGPT session in Chrome, a compatible Codex/browser bridge exposing globalThis.agent, and specific Chrome extension ("Allow access to file URLs") and Codex app ("Allow Google Chrome uploads") permissions for file operations.
  • Links: GitHub Repository

Highlighted Details

  • Operates exclusively through visible ChatGPT sessions via a browser bridge.
  • Provides workflow primitives, not a direct OpenAI API replacement.
  • Integrates seamlessly as a Codex plugin.
  • Supports local file uploads and image artifact downloads (requires specific permissions).
  • Generates local run reports omitting sensitive prompt/response content by default.

Maintenance & Community

No specific details on contributors, sponsorships, community channels (e.g., Discord, Slack), or a public roadmap are provided in the README.

Licensing & Compatibility

The README does not specify a license. The project is explicitly not affiliated with, endorsed by, or sponsored by OpenAI. Safety guidelines suggest restrictions on use cases that bypass user-directed visible sessions or violate terms of service.

Limitations & Caveats

This is an unofficial alpha SDK. It strictly requires a visible ChatGPT session and a compatible browser bridge; ordinary shell execution without a bridge will result in a browser_bridge_unavailable error. It is not a general-purpose browser automation tool. File upload functionality is gated by specific browser and application permissions.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
15
Issues (30d)
0
Star History
122 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.