Navari

Navari is an agentic AI platform for R&D scientists. You describe the work and the agents do it: a literature review across PubMed and the clinical trial registries, a data analysis, a slide deck, a working application. Each capability is an MCP server, so adding a data source or a tool never touches the reasoning core.

Where it came from

It started small. A few of us were building agents and workflows to get our own work done faster and sharing what worked. I founded it and set the product direction; an engineering lead built the first version of the framework in about six weeks. It launched to a pilot group in spring 2025. Nobody marketed it. A scientist would finish in an afternoon what used to take a week, a colleague would ask how, and the colleague would be on the platform the next day.

How it's built

• Model-agnostic by design. Claude was the first reasoning engine; the platform swaps backends without a rebuild.
• Tool execution through MCP: literature, clinical trial registries, regulatory data, document and slide generation, data analysis, enterprise data.
• React frontend, Python backend, AWS, enterprise auth and audit logging, with every session isolated per user.
• Governed, traceable outputs, because an R&D decision has to show its sources.

The numbers

• 1,000+ active users while I ran it, with 2,900% year-on-year growth at the peak.
• 10+ MCP servers in production by the end of 2025, more since.
• Handed off in mid-2026. As of late 2026 it's reported at about 10,000 users across the organization. That number is second-hand, which is awesome in its own way: it kept growing without me.

Where it fits

Navari is the execution layer of a three-part stack. Agentis decides which clinical questions deserve an agent and why. AIthena makes the trial data AI-ready so the agents have something clean to work on. Navari builds and runs the agents. ARIA is the mirror image: Navari puts agents to work for a scientist on demand, ARIA runs research unattended for months. The pharma chapter of ARIA was pointed at the questions a growing agent platform raises.