Enterprise-Ready MCP // Jiquan Ngiam

Why It Matters

Enterprise AI agents are reshaping software development, but unchecked deployment can expose critical data and operational stability. Robust MCP governance ensures organizations reap productivity gains while maintaining security and compliance.

Key Takeaways

• 80% developers use AI tools daily
• Agentic coding platforms like Claude Code surge
• MCP gateways provide secure AI‑data connections
• Guardrails essential to mitigate agent security risks
• Prosus backs production‑grade AI agent conference

Pulse Analysis

Enterprise adoption of AI‑driven agents is accelerating faster than any previous technology wave. Recent surveys show more than 80 % of professional developers invoke AI tools on a daily basis, and platforms that embed code‑generation agents—such as Anthropic’s Claude Code—are reporting double‑digit growth in active users. This surge is driven by the promise of Model Context Protocols (MCPs), which let developers attach rich, domain‑specific data to large language models, turning generic AI into a task‑specific assistant that can read, write, and execute code within a company’s own environment.

The rapid rollout, however, surfaces a new class of security concerns. When an agent can query internal databases or trigger deployments, malicious prompts or unintended behaviors can cause data leakage, compliance breaches, or production outages. Experts like Jiquan Ngiam advocate for hardened MCP gateways that enforce authentication, audit trails, and policy‑based throttling, effectively sandboxing the AI’s reach. Complementary guardrails—static prompt validation, runtime monitoring, and automated rollback—are becoming mandatory components of any enterprise AI stack, ensuring that the convenience of autonomous coding does not outweigh risk.

Investors and cloud providers are taking notice, with Prosus Group sponsoring the upcoming Coding Agents Conference to showcase production‑grade solutions. The event signals a maturing ecosystem where MLOps platforms, security vendors, and AI model providers converge on standardized protocols. As enterprises embed agents deeper into CI/CD pipelines, the market for secure MCP infrastructure is poised to expand, creating opportunities for startups like MintMCP and prompting larger players to integrate comparable governance layers into their AI services.

Original Description

March 3rd, Computer History Museum CODING AGENTS CONFERENCE, come join us while there are still tickets left.

https://luma.com/codingagents

Thanks to @ProsusGroup for collaborating on the Agents in Production Virtual Conference 2025.

Abstract //

Agents and Model Context Protocols (MCPs) are being rapidly adopted across enterprises: over 80% of professional developers now using AI tools daily, and agentic coding platforms like Claude Code seeing significant growth. This session will explore emerging patterns and security risks, along with strategies to mitigate them. We’ll share insights from observing agent–MCP interactions, discuss methods to detect and prevent potentially harmful behaviors, and outline practical approaches for establishing robust guardrails to ensure safe and controlled MCP deployment.

Bio //

Jiquan Ngiam works on solutions that connect AI to data and applications. He is the co-founder and CEO of MintMCP, a company building enterprise AI infra including MCP gateways that securely connects AI assistants to internal company data and tools with enterprise-grade governance, auditability, and security. Previously, he was a senior staff researcher at Google Brain, on the founding team at Coursera, and in Andrew Ng’s Stanford lab, where he co-authored pioneering work in multimodal deep learning. His technical work spans bridging AI models to real systems, and he is now focused on building reliable, enterprise-ready AI infrastructure.

A Prosus | MLOps Community Production