robertpiosik/CodeWebChat

What it solves

CodeWebChat provides a privacy-focused way for developers to use AI for pair programming without being locked into a single AI provider or expensive subscription. It solves the friction of manually copying and pasting relevant code files into web-based AI chatbots by automating the construction of a "static context" prompt.

How it works

The toolkit consists of a VS Code extension and a companion browser extension. The editor extension allows users to select specific files, search for relevant code, or identify changed files in Git. It then bundles these files into a structured Markdown prompt. Users can either send this prompt directly to an AI via an API key or use the browser extension to "autofill" the prompt into various web-based AI interfaces (like ChatGPT, Claude, or Gemini) without manual copying.

Who it’s for

Developers who want to use state-of-the-art reasoning models for coding but prefer to maintain full control over their data, use their own API keys, or leverage the free web interfaces of various AI providers.

Highlights

• Static Context Approach: Sends all relevant files in one unified message to avoid multiple tool calls and increase accuracy.
• Privacy-First: Operates entirely on the local machine; no code or prompts are collected.
• Broad Compatibility: Supports a wide array of web chatbots including AI Studio, Claude, ChatGPT, DeepSeek, and OpenRouter.
• Prompt Caching Optimization: Orders files by recency and places instructions at the end to maximize the efficiency of prompt caching.
• Integrated Git Workflow: Includes commands to select unstaged files or files from specific commits to generate targeted prompts.