
Cococure
Hospitality, Nightlife
An AI concierge that handles restaurant and nightclub bookings over WhatsApp and voice: it checks availability, makes the booking and recommends venues through MCP server tools.
An AI agent layer for a diagnostics network whose patient app has 1M+ downloads: plain-language lab-report explanations, test guidance, lab lookup and voice, orchestrated across MCP tool servers.
Patients struggle to understand their lab reports, which test to take next and where to go, while the lab's support staff answer the same questions all day.
Architected an orchestrator agent in PydanticAI that routes each request to specialised MCP tool servers for tests, reports, radiology documents, locations and general information.
The assistant explains lab reports in plain language, suggests relevant tests and answers questions about the lab's services.
Location tools find the nearest collection centres.
Added voice through VAPI, so patients can speak to the assistant as well as type.
Routed models through OpenRouter, including GLM.
Built evaluation sets from sample reports and patient-style questions, and ran them after every iteration and in production.
Traced every agent run and tool call in Logfire.
Kept each capability in its own MCP server, so tools can be tested, updated and deployed independently.
Orchestrator agent
Specialised MCP tool servers
Lab-report explanation and test guidance
Location lookup
Voice interface
Logfire observability
We can scope and build a system with comparable capabilities, shaped around your own workflows, data and integrations.