Introduction to OpenCrow

OpenCrow is an open-source AI agent platform designed to embed intelligent assistants into web applications. It allows developers to create, manage, and deploy AI agents that can interact with both server-side APIs and client-side application logic.

What OpenCrow Can Do

• Embeddable Widget: Provide a React-based chat widget to your users with a single line of code or custom component installation.
• Split-View Dashboard: Manage agents, sessions, and configurations easily from an admin dashboard.
• Dual-Layer Tooling: The agent can interact with your backend server through OpenAPI specifications AND manipulate the client's frontend (like navigating pages, opening sidebars).

What OpenCrow Cannot Do (Yet)

• Vector Knowledge Base parsing is under development and not fully supported for all file types yet.
• Advanced custom authentication flows within the widget are partially supported and require custom implementations.

Cloning and Setup

OpenCrow is designed as a monorepo. To get started locally:

# Clone the repository
git clone https://github.com/himasnhu-at/opencrow.git
cd opencrow

# Install dependencies (requires pnpm)
npm install -g pnpm
pnpm install

Environment Configuration

You'll need to set up .env files in two locations. Check the .env.example files in each directory:

1. Dashboard: apps/app/.env
2. Backend: apps/backend/.env (Requires OpenAI/Gemini API keys, Database URLs, etc.)

Running the Project

Start the required backend services (PostgreSQL, Redis, etc.) using Docker:

docker-compose up -d

Once the containers are running, start all development servers:

pnpm dev

Access Points

• Dashboard: http://localhost:3000
• Backend API: http://localhost:3001
• Landing Page: http://localhost:4000
• Sample Store (Demo): http://localhost:8000

Integrating the Widget

To add the widget to your Next.js/React site, first install the package (assuming you're in the monorepo):

import { OpenCrowWidget } from "@opencrow/widget";

export default function MyPage() {
  return (
    <>
      <h1>Welcome to My Site</h1>

      <OpenCrowWidget
        projectId="your-project-id"
        tools={{
          navigate_to_page: ({ page }) => {
            window.location.href = `/${page}`;
            return { success: true };
          },
          open_cart: () => {
            // your logic here
            return { success: true };
          },
        }}
      />
    </>
  );
}

Define any custom tools you want the AI to be able to execute directly on the frontend.