Botsify Agentic
AI Agent platform — frontend, realtime, MCPs, white-label
About this project
I handled the complete frontend and backend modules: realtime conversation, analytics, building agents, MCPs, and third-party integrations (WhatsApp, Facebook, Meta). Key challenge: agent prompts — we built thinking, building, and suggestion steps plus a final prompt that runs on the chatbot. We rewamped from a workflow-based system to a conversation-based one; I focused on the agents page, implemented OpenAI response API, MCPs, and web/file search. The hard part was system_prompt design so the agent could replicate what customers do with drag-and-drop. Fixed a critical bug where connecting the same MCP twice (e.g. two Google Sheets) broke the bot — MCP wasn’t creating unique sessions; fixed after hours of debugging. For white-label we use a Cloudflare Worker so whitelabel.com/api routes to our API without exposing our backend URL.
Key features
- Real-time Chat & Streaming
- MCP, Web & File Search
- Agent Builder & System Prompts
- White-label + Proxy (Cloudflare)
- WhatsApp, Facebook, Meta
- Analytics & Unique MCP Sessions