Weekly AI insights —
Real strategies, no fluff. Unsubscribe anytime.
A community-focused startup needed a platform combining live video conferencing, instant messaging, event scheduling, and a complex permission system — essentially a modern alternative to Slack and Discord with built-in video rooms. The technical challenge was real-time state synchronization across hundreds of concurrent users with sub-second latency, while maintaining a permission model that operates at organization, channel, and room levels..
4 weeks
Development Time
Full platform with video
<200ms
Message Latency
Real-time propagation
30+
Pages Built
Including admin console
25+
API Endpoints
All with permission checks
9
Responsive Breakpoints
320px to 4K tested
100%
Permission Checks
Every endpoint secured
The Challenge
A community-focused startup needed a platform combining live video conferencing, instant messaging, event scheduling, and a complex permission system — essentially a modern alternative to Slack and Discord with built-in video rooms. The technical challenge was real-time state synchronization across hundreds of concurrent users with sub-second latency, while maintaining a permission model that operates at organization, channel, and room levels.
Discovery Phase
We mapped the permission model first — this was the architectural foundation everything else depended on. We identified 5 permission levels (owner, admin, moderator, member, guest) across 3 scopes (organization, channel, room). We then documented every real-time interaction: message sending, typing indicators, presence status, video room join/leave, and screen sharing. The critical constraint was that every state change needed to propagate to all connected clients within 200ms.
Our Solution
We chose Convex for real-time data synchronization — its reactive query model means UI components automatically re-render when underlying data changes, eliminating manual WebSocket management. LiveKit handles the video infrastructure with SFU (Selective Forwarding Unit) architecture for low-latency multi-party video. Clerk provides authentication with organization management, and we built a custom permission middleware that checks access at every API endpoint.
Architecture
The application uses a reactive architecture where every database query is a real-time subscription. When a user sends a message, all clients viewing that channel see it instantly through Convex reactive queries — no polling, no manual WebSocket management. The video layer is decoupled from the messaging layer, allowing independent scaling. LiveKit rooms are created on-demand and destroyed when empty, optimizing infrastructure costs.
Technology Stack
Key Results
LiveKit video integration for live calls with screen sharing and recording
Real-time messaging with Convex reactive queries and sub-200ms propagation
5-level role-based access control across organizations, channels, and rooms
Event scheduling with calendar integration, reminders, and attendance tracking
Community analytics dashboard with engagement metrics and activity heatmaps
Mobile-responsive design tested across 9 breakpoints (320px to 4K)
Typing indicators, presence status, and read receipts — all real-time
File sharing with preview, drag-and-drop upload, and storage management
Features
Core features delivered for Kommu.
Multi-party video conferencing with screen sharing
Real-time messaging with threading and reactions
Organization and channel management
5-level permission system (owner to guest)
Event scheduling with calendar sync
Community analytics and engagement metrics
File sharing with drag-and-drop
“The speed is unreal. Full SaaS with real-time messaging, video rooms, and Stripe billing — production-ready in under a month. We thought it would take a team of 8 at least 4 months. The real-time performance is indistinguishable from native apps.”
Lessons Learned
Design the permission model before writing any application code — it touches everything
Convex reactive queries eliminate an entire category of real-time sync bugs
Video infrastructure should be decoupled from the application layer for independent scaling
Full-stack developer and AI architect with years of experience shipping production applications across SaaS, mobile, and enterprise. Gareth built Agentik {OS} to prove that one person with the right AI system can outperform an entire traditional development team. He has personally architected and shipped 7+ production applications using AI-first workflows.
More Work
Whether it is a SaaS, mobile app, or enterprise platform — we build it in weeks, not months.