Kelvin Yu
Contributing Writer, Explore Agentic
About Kelvin
Kelvin covers the AWS-native agent platform and the control planes wrapped around it. On the runtime side that means Bedrock, AgentCore Runtime and the AgentCore Gateway that went GA on October 13, 2025, Amazon S3 Vectors weighed against OpenSearch Serverless for enterprise RAG, and a shipping comparison of AWS AgentCore against Azure AI Foundry drawn from running MCP servers and agents on both clouds.
On the control-plane side Kelvin writes the agent registry and agent gateway field guides plus the MCP gateway glossary entry: what a registry has to record about every agent and MCP server in the estate, what a gateway must refuse at runtime, why A2A moving from v0.3 to v1.0 in March 2026 invalidated assumptions teams had already built on, and which OAuth extensions enterprise IdPs never shipped for agents — protected resource discovery, dynamic client registration, long-lived machine tokens, on-behalf-of delegation. The observability entry maps OpenTelemetry and W3C trace context onto non-deterministic model calls. Expect deployment detail that survives a security review.
Pieces written or reviewed by Kelvin
22 pieces across 3 formats on this site — 12 written by Kelvin and 10 reviewed. A written byline means Kelvin researched and drafted the piece; a reviewed byline means Kelvin read it against its cited sources and could defend its claims before it published. Every row below is labelled either way.
- Glossary · Written MCP Gateway
- Glossary · Written Agent observability
- Comparison · Written Glean vs ChatGPT vs Claude on enterprise search: auditing the '1.9× preferred' eval claim
- Insight · Reviewed AWS Bedrock Pricing for Agent Builders: What Your Agents Actually Cost in 2026
- Insight · Written Amazon S3 Vectors for Enterprise RAG: When the Cheapest Vector Store Is the Right One
- Insight · Reviewed SageMaker vs Bedrock in 2026: The Decision Framework for Agentic Workloads
- Insight · Reviewed LLM Observability Isn't Optional: What Technical Leaders Need to Build It Right
- Insight · Reviewed Prompt Versioning Done Right: Treating Prompts as Deployable Artifacts
- Insight · Written Claude for Enterprise: The Three Buying Paths (Anthropic API, Amazon Bedrock, Team/Enterprise Seats) and What Each Really Costs
- Insight · Reviewed Claude on Amazon Bedrock: The Token Economics of Agentic Workloads
- Insight · Reviewed The Claude Discount Nobody Quotes: How Enterprises Actually Cut Claude Spend
- Insight · Written Glean Skills and Adaptive Reasoning: is the "enterprise AI coworker" an agent runtime, a workflow runner, or a relabel?
- Insight · Written Glean year-zero cost: what a 500-seat deployment actually burns before the first renewal
- Insight · Written AWS AgentCore vs Azure AI Foundry: Lessons from Shipping MCP Servers and Agents on Both Clouds
- Insight · Written Enterprise AI Without the Security Compromise: A Framework for Governed AI Deployment
- Insight · Written Glean renewal checklist: five signals to renegotiate or leave before signing year two
- Insight · Reviewed When MCP Gateway Auth Breaks: Practical Patterns for a Developer-Friendly Experience
- Insight · Reviewed MCP Server Security: The Threat Model and a Hardening Checklist for Tool-Calling Agents
- Insight · Written Multi-Agent Orchestration Patterns: Supervisor, Swarm, and Pipeline
- Insight · Reviewed AI Agents for Insurance Eligibility Verification: What Clinics Can Automate Today — and What Still Needs a Human
- Insight · Reviewed Claude Team and Enterprise Seat Pricing in 2026: What a Seat Costs and What You'll Actually Spend
- Insight · Written Agent Gateway vs. API Gateway: What Your Existing Gateway Cannot Enforce
Other contributors
- Ryo HangEditor-in-Chief
- Alexander GromanContributing Writer
- Soraya ZhengContributing Writer
- Cynthia ZhangContributing Writer
- Tommy TaoContributing Writer
- Chandler BensonContributing Writer
- Elias SaljukiContributing Editor
- Ginny CasuccioContributing Editor
- Gloria Qian ZhangContributing Editor
- Laura Bradley McCoyContributing Writer
- Merve TengizContributing Editor
- Michael CloughEditorial Advisor