Obot

AI assistant browser extension and backend built with Cloudflare Workers, Durable Objects, and React.

Obot is a full-stack AI assistant consisting of a Chrome Extension Side Panel (frontend) and a Cloudflare Workers backend. It uses Cloudflare Durable Objects, Workers AI, and the Model Context Protocol (MCP) to provide real-time chat and tool-calling capabilities.

---

📦 Latest Extension Release

You can download pre-built, ready-to-install extension bundles directly from GitHub Releases:

• Download the obot-extension-v*.zip asset from the latest release.
• Unzip/extract the file on your computer.
• Open chrome://extensions, enable Developer mode, and click Load unpacked.

✨ Key Features

• Chrome Extension: React-based panel that lets you chat, ask questions about your active tab, and summarize page content.
• Model Context Protocol (MCP): Integrates with external APIs (such as Exa Search, Mem0, Apify, and custom self-hosted MCP servers) using the MCP standard.
• Durable Objects Architecture: Manages stateful chat sessions as Durable Objects over persistent WebSocket connections.
• Cloudflare Workers AI: Uses open-source LLMs (Llama 3.2, Gemma 2B, Qwen, GLM, GPT-OSS) hosted on Cloudflare's platform.
• Dynamic Code Execution: Uses @cloudflare/codemode to generate and run JavaScript that orchestrates MCP plugin actions.
• OAuth 2.0: Uses Google Identity to authenticate users and persist configuration state across sessions.

🚀 Quick Start

This repository is set up as a monorepo containing both the extension and worker backend.

Prerequisites

• Node.js (v18 or higher recommended)
• Wrangler CLI (for Cloudflare deployment)
• A Cloudflare account with D1 and KV enabled

1. Backend Setup (Worker)

Navigate to the worker/ directory and configure the environment:

cd worker
cp .dev.vars.example .dev.vars

Add your credentials (such as Google OAuth keys, AI provider secrets, etc.) to .dev.vars.

Run D1 database migrations to initialize the schema:

npm run migrate

Start the local worker dev server:

npm run dev

2. Frontend Setup (Extension)

For local development (runs the WXT dev server, loads the unpacked extension automatically, and connects to http://127.0.0.1:8787):

npm run dev:ext

To build the extension for production (default worker URL https://api.linkos.in; override with WXT_WORKER_URL in extension/.env):

npm run build:ext

The build output is generated into extension/.output/chrome-mv3/.

3. Load the Extension in Chrome

1. Open Google Chrome and navigate to chrome://extensions/.
2. Toggle the Developer mode switch in the top-right corner.
3. Click Load unpacked in the top-left corner.
4. Select the extension/.output/chrome-mv3/ directory generated by npm run build:ext.
5. Click the Extensions puzzle icon in the toolbar, pin Obot, and click it to open the side panel.

---

💻 Workspace Command Reference

You can run these scripts from the root directory to orchestrate the entire project:

Command	Action
npm run dev:worker	Starts wrangler local development server for the backend (http://127.0.0.1:8787)
npm run dev:ext	Runs the WXT dev server for the Chrome extension side panel (auto-loads unpacked)
npm run build:ext	Builds the Chrome extension for production (https://api.linkos.in)
npm run build:worker	Deploys the worker backend directly to Cloudflare
npm run migrate	Runs the initial D1 schema migrations for the database

---

🛡️ Architecture & Tech Stack

Frontend Stack

• React 19 & Lucide Icons for the side panel UI.
• WXT for the extension scaffold and asset bundling.
• agents SDK & @cloudflare/ai-chat to manage persistent WebSocket connections directly to Cloudflare Durable Objects.

Backend Stack

• Hono Framework for routing, OAuth flow, and API endpoints.
• Durable Objects (ChatAgent) to host stateful, persistent chat threads and manage active MCP plugins per-user.
• Workers AI & D1 (SQLite) for database operations, session storage, and LLM inference.