Turning Apify into a Production MCP Gateway
A Real-Time Voice- and Tool-Driven Transport Assistant
Aliyu Abdulbasit
Funsho Olalere
Presented by
What we builit
Company Name
Quarter Month
Year
1
2
3
4
5
6
Understand Czech Transport & Air Quality
Calls real MCP tools for live data
Speaks back instantly using Elevenlabs TTS
Can self-monitor delays using n8n
Runs fully end-to-end inside Docker
Add section title
+
The Problem
People waste time searching for Transport apps for departures, delays and routing. → Visually impaired users struggle even more.
We built a hands-free agent that replaces:
The Solution
Speech-to-text (Whisper)
LLM reasoning (GPT-4o-mini)
MCP tools from CzechFabric
n8n automation
TTS from ElevenLabs
Chainlit UI
Key Features
Company Name
Quarter Month
Year
1
2
3
4
5
6
Real-time WebSocket voice interface
Autoplay ElevenLabs audio replies
Live departures, route planning, air quality
Autonomous delay alerts via n8n workflows
Runs fully end-to-end inside Docker
Bilingual: Czech & English, TONE format → 40 percent fewer LLM tokens
+
End-to-End Flow
The user speaks/types → VAD → Whisper STT → LLM → MCP tools → LLM → ElevenLabs → audio stream back.
Short Demo Video
Why It’s Innovative
We integrated 12+ live Czech government APIs through CzechFabric MCP:
Designed for pure tool reasoning.
Real-World Use Cases
Thank you
Quarter Month
Year
Company Name
We built CIA to show the future of voice agents in public infrastructure
.
Links:
• GitHub: https://github.com/rightoncue-ai/elevenlabs_ai
• MCP Server: https://czechfabric.cz