A proxy between your code agent and your real browser.
bproxy lets coding agents read pages, fill forms, and click through tasks inside your signed-in Chrome session — without exposing an automation handle to the page. A Chrome extension executes a narrow set of commands; a localhost daemon paces and scopes them; a CLI gives the agent a one-shot interface. Login, CAPTCHA, consent screens, and final submits stay yours.
The motivation, design principles, scenarios, and architecture live in the documentation site. The README is a quick orientation — the docs are the source of truth.
Documentation: https://dimdasci.github.io/bproxy/
Code Agent ──CLI──▶ Proxy Daemon ◀──WebSocket──▶ Browser Extension ◀──▶ Page
Install the CLI and daemon from npm:
npm install -g @dimdasci/bproxyThe Chrome extension is currently a manual unpacked install pending a Chrome Web Store listing — download the matching zip from GitHub Releases and load the extracted folder via chrome://extensions with Developer mode on.
The Install guide walks through both steps end to end. Usage is the CLI command reference, and Upgrade covers version bumps.
bproxy is a TypeScript monorepo with three runtime workspaces (cli, service, extension) and a shared protocol package (shared). The contracts that govern how they fit together — actions, errors, sessions, tabs, write methods — are documented in the Solution Specs, and the system shape is captured in the Architecture views. Internal artifacts (ADRs, phase plans, journal entries) live under docs/internal/ in this repository.