Agentic AI Tachles
"Agentic AI Tachles" is a technical meetup group focused on the practical, "no-fluff" side of building agentic AI systems. Designed for engineers and builders who learn by making, the group explores the intersection of AI innovation, security, and infrastructure.
Sessions cover critical real-world challenges, including:
Agent Security: Identifying and mitigating vulnerabilities in MCP (Model Context Protocol) and agentic workflows.
System Architecture: Moving beyond "happy path" demos to build responsive, composable multi-agent systems using protocols like A2A and SSE.
The Builder's Mindset: A hands-on approach to problem-solving that prioritizes reliability, infrastructure efficiency, and open-source primitives.
Whether you are designing agent orchestration, managing production-scale infrastructure, or simply curious about what it takes to move AI agents from prototype to production, this group provides a space to share deep technical insights and practical experience.
Sessions
MCP and the Chamber of Secrets / Neta Ravid, Agentic Security Lead, Astrix
As identity security engineers, we use AI for productivity while solving its security flaws. We discovered a major risk: 60% of MCP servers rely on plaintext, static credentials. This talk explores how MCP setups leak tokens and presents our mitigations, including secret vaulting and open-source tools.
The Builder's Mindset in the AI Era / Nir Adler, AI Innovation Engineer
Understand problems by building solutions. We'll explore a stack of open-source projects—from AI agents with "eyes" to voice APIs and centralized orchestration. We also address scaling employee agent bots using Linux namespaces for safe, cheap sandboxing. Whether hardware or software, the goal is to build small and bridge gaps.
Beyond the Slop: Real-Time Agent Communication with A2A and SSE / Dima Groisman, Principal Engineer, Firefly
Move past flashy demos to resilient systems. Learn how the Agent-to-Agent (A2A) protocol enables delegation and coordination, while Server-Sent Events (SSE) provide real-time visibility. This code-forward session covers architecture and gotchas for building composable, responsive agent systems.