Open Source Summit India 2026

AI agents are often limited by proprietary "plugin" silos and bespoke integrations. The Model Context Protocol (MCP) offers a universal interface to break these walls, but moving from local scripts to enterprise-scale introduces Day 2 operational hurdles. How do we discover tool-servers in a dynamic cluster? How do we secure the bridge between LLM reasoning and system execution?

This session presents a technical blueprint for operationalizing MCP within a sovereign, open source stack. We move beyond basic connectivity to address 3 core pillars:
1) Dynamic Discovery: Implementing cloud-native service discovery so agents can dynamically negotiate capabilities with ephemeral MCP servers.
2) Zero-Trust Security: Hardening the "Reasoning-to-Execution" path with fine-grained authorization and audit trails to prevent privilege escalation.
3) Full-Stack Observability: Using OpenTelemetry to bridge the gap between model thinking and backend execution, pinpointing failures in production pipelines.

Attendees will leave with a roadmap for building a modular, interoperable agentic infrastructure, ensuring the future of AI remains open and auditable.