BridgeCalle

Frontpage
LogsSummary

💡 Inspiration

Over 70 Million seniors worldwide suffer from chronic loneliness and social isolation. While family members care deeply about their aging parents, busy work schedules make long daily phone catch-ups difficult.

Existing AI voice bots fail seniors because they talk too much—interrupting, offering unsolicited advice, or lecturing.

We created BridgeCalle to flip the script: an active-listening voice companion powered by CALL-E that listens 95% of the time, letting seniors speak freely, and automatically generates structured call summaries that family members can share directly to WhatsApp.

⚙️ What it does

🖼️ Illiterate-Friendly 1-Tap Photo Dialing: Seniors don't have to navigate complex phone menus. They simply tap a familiar picture or name on any smartphone, tablet, or PC to immediately receive a CALL-E AI phone call.
👂 95/5 Active Listener Persona: The AI greets gently with a water and meal check-in ("Did you take your water and ate food today?"), then stays quiet, offering only short 1-2 word verbal nods ("Mmhmm", "How great") so the senior feels heard.
📋 Structured Call Summaries: Tracks every call sequentially (Call #1, Call #2, etc.) with exact start time, duration, and bulleted key points.
💬 1-Tap WhatsApp Summary Export: Allows family members to export call digests directly to WhatsApp chats so they can stay updated and call back when free.

🛠️ How we built it

BridgeCalle utilizes all 5 CALL-E modalities:

CALL-E REST API (POST /v1/calls): Real-time outbound call task dispatching with task prompts and regional locale settings (en-IN / en-US).
CALL-E SDK (@call-e/calle): Server-side call task orchestration.
CALL-E MCP (bridgecalle-mcp.json): Streamable HTTP Model Context Protocol integration for AI agents.
CALL-E CLI (src/cli/call-e-cli.js): Command-line utility to trigger active-listening phone calls directly from terminal.
CALL-E Agent SKILL (skills/bridgecalle-active-listener/): Packaged reusable Agent Skill with SKILL.md and safety/example guidelines.

The frontend is an ultra-accessible, responsive web app deployed live on GitHub Pages, built with HTML5, Tailwind CSS, and local storage persistence.

🚧 Challenges we overcame

Enforcing AI Silence: Standard LLMs tend to over-explain and interrupt. We engineered CALL-E's task instructions to strictly enforce a 95% listening ratio with 3,500ms patient silence thresholds so seniors are never rushed.
Connection Speed: Optimized the task prompt into a direct, fast instruction so CALL-E connects the phone call within seconds without long execution planning delays.
Non-Literate Accessibility: Designed a photo-first grid layout where tapping anywhere on a familiar picture triggers a call instantly.

🎓 What we learned

How to leverage CALL-E's goal-driven execution model to automate real phone calls that adapt in real time.
Designing zero-friction interfaces tailored specifically for elderly and non-tech-savvy users.

🔮 What's next for BridgeCalle

Adding multi-lingual active listening support for Gujarati, and regional Indian languages.
Integration with smart home hardware and landline automated dialers for seniors without smartphones.