Here’s what made it stand out:
Built for real-time AI (not just calls) – LiveKit handles audio, video, and data streams together, making it ideal for multimodal AI agents
Low latency, high performance – powered by WebRTC + SFU architecture for smooth, scalable real-time communication
Open-source & flexible – we can customize, self-host, and deeply integrate into our stack instead of being locked into rigid workflows
Developer-first architecture – clean SDKs across platforms and faster iteration for complex use cases
Handles infra complexity – abstracts the hardest parts of WebRTC while still giving us control
erplexity stood out for a few key reasons:
Source-backed answers by default – every response comes with citations, making it easy to verify and trust the output
Real-time web intelligence – unlike most models relying on static training data, it fetches live information from the internet
Faster research workflows – instead of scanning multiple links, it gives a direct, consolidated answer in one go
Built for depth, not just conversation – excels at structured research, summaries, and fact-checking across multiple sources
Clean, unbiased experience – ad-free and focused on accuracy rather than engagement loops
Kipps AI gave us:
1. Multi-modal capabilities (text, image, video, voice) in one stack, not just voice pipelines
2. Better control over agent behaviour with customizable workflows, memory, and persona tuning
3. Faster iteration cycles for launching and refining AI features inside ModeAI
4. Unified infrastructure instead of stitching together multiple tools
Kipps AI WhatsApp Agent runs on secure, reliable cloud infrastructure powered by Amazon Web Services, helping us deliver fast, scalable AI experiences to businesses worldwide.
Vapi
