Architecture Overview

Relevant source files

The following files were used as context for generating this wiki page:

• README.md
• middleware.ts
• app/page.tsx
• app/api/task/route.ts
• app/api/prs/route.ts
• app/api/review/route.ts
• lib/auth.ts
• package.json

GeNCorE (Claude Remote Control) is a command center designed to manage AI agents such as GitHub Copilot, Claude, and Codex. It provides a centralized web interface to delegate tasks, review pull requests, and monitor agent status by interacting directly with the GitHub REST API.

Sources: README.md:1-5, package.json:2-4

System Components

The project is built using Next.js 15 with the App Router architecture. It is divided into a client-side frontend for user interaction and a server-side API layer that handles secure communication with external services.

Frontend Layers

The frontend is built with React and Tailwind CSS, utilizing a mobile-first design optimized for devices like the iPhone. Key views include:

• Dashboard (/): Displays agent status and a list of active Pull Requests.
• Task Panel (/task): A form to assign new tasks to specific AI agents by creating GitHub Issues.
• PR Review (/review): An interface for approving, merging, or requesting changes on Pull Requests.

Sources: README.md:7-14, app/page.tsx:75-156, app/task/page.tsx:64-135

Backend API Routes

The backend consists of Next.js Route Handlers that proxy requests to the GitHub API using a server-side GITHUB_TOKEN. This ensures sensitive credentials never reach the client.

Endpoint	Method	Description
/api/prs	GET	Fetches open PRs from the configured GitHub repository.
/api/task	POST	Creates a new GitHub Issue assigned to an agent.
/api/merge	POST	Merges a specific PR using squash, merge, or rebase methods.
/api/review	POST	Submits a review (APPROVE/REQUEST_CHANGES) to a PR.
/api/chat	POST	Direct interface for interacting with the Anthropic Claude API.

Sources: README.md:18-24, app/api/prs/route.ts:9-47, app/api/task/route.ts:10-77

Data Flow & Integration

The system acts as an intermediary between the user and GitHub. Below is a visual representation of how a task is delegated through the system.

Pull Request Management

The application identifies which agent is responsible for a PR by detecting patterns in the branch name (head.ref) or the user login.

Security Architecture

Security is implemented through a combination of middleware, environment variables, and cryptographic hashing.

Authentication Flow

The application is protected by an ADMIN_TOKEN. Access is managed via a session cookie containing an HMAC-SHA256 hash of the token, preventing the raw password from being stored in the browser.

1. Hasing: The hashAdminToken function uses the Web Crypto API to sign a static payload with the ADMIN_TOKEN.
2. Middleware: Every request (except /login and /api/auth) is intercepted by middleware.ts.
3. Timing Protection: Verification uses timingSafeEqual to prevent timing attacks during hash comparison.

Sources: lib/auth.ts:13-41, middleware.ts:25-63

Validation and Protection

• Input Validation: PR numbers are validated as safe integers (1 to 2,147,483,647) to prevent SSRF (Server-Side Request Forgery) when constructing GitHub API URLs.
• Credential Isolation: GITHUB_TOKEN is strictly server-side.
• Production Safety: If ADMIN_TOKEN is missing in a production environment, the middleware forces a 500 error to prevent public access.

Sources: README.md:92-100, app/api/review/route.ts:31-35, middleware.ts:38-51

Technical Stack Summary

Component	Technology
Framework	Next.js 15 (App Router)
Language	TypeScript
Styling	Tailwind CSS
AI SDK	@anthropic-ai/sdk
Deployment	Vercel
External API	GitHub REST API v3

Sources: package.json:15-30, README.md:77-83

Conclusion

The GeNCorE Architecture leverages a stateless API proxy model to provide a secure, mobile-optimized command center for AI agents. By integrating tightly with GitHub's Issue and Pull Request systems and securing the boundary with HMAC-based session management, it allows for safe remote control of automated development workflows.