Home›MCP

MCP

Model Context Protocol (MCP) is an open standard for connecting AI assistants to tools, data sources, and APIs. A rapidly growing pattern for building composable agentic systems.

25 articles

MCP deep dives from the blog

All posts →
MCP

Model Context Protocol (MCP) Explained for Engineers: Protocol Design, Function Calling Trade-offs, and Building Your First Server

13 min read · Read →
AI Agents

How Agent Skills Are Reshaping Agentic Workflows in 2026

12 min read · Read →
LLM Engineering

Tool-Calling and Agent Sandboxing Security in Production

13 min read · Read →
25 articles
Govern and discover records with common metadata in AWS Agent Registry
AWS ML Blog· 12 min read· Yesterday
Govern and discover records with common metadata in AWS Agent Registry

AWS Agent Registry now supports custom metadata schemas, letting administrators define organization‑wide key‑value attributes (e.g., business unit, risk classification) that every MCP server, skill, or agent record must adhere to. Records are validated against the schema at creation or update, yielding a COMPLIANT or…

How to Design MCP Servers That Leave Orchestration to the Host
Towards Data Science· Yesterday
How to Design MCP Servers That Leave Orchestration to the Host

The article presents a Python SDK v2 catalogue server that implements a Model Context Protocol (MCP) server exposing two narrow tools, a single resource, and one prompt, deliberately keeping planning, workflow state, and authorisation on the host side. This design demonstrates how MCP servers can offload orchestration…

Automating AWS Agent Registry updates with push-based architecture
AWS ML Blog· 17 min read· Nov 11, 2026
Automating AWS Agent Registry updates with push-based architecture

The article details a fully event‑driven pipeline that keeps the AWS Agent Registry up‑to‑date by reacting to MCP server changes. It wires AWS CloudTrail to Amazon EventBridge, which triggers an AWS Lambda function that uses AgentCore Identity to push new tool metadata to the registry without storing any secrets. The…

Serverless MCP server for AI agent data lake access on Amazon ECS Fargate
AWS ML Blog· 14 min read· Nov 6, 2026
Serverless MCP server for AI agent data lake access on Amazon ECS Fargate

The authors demonstrate how to spin up a fully serverless Model Context Protocol (MCP) server on Amazon ECS Fargate that gives AI agents secure, metadata‑driven access to an enterprise data lake. The approach eliminates per‑dataset integration code by requiring only a JSON descriptor to expose a new dataset,…

Workflow context for AI agents with Smartsheet MCP on AWS
AWS ML Blog· 10 min read· Oct 20, 2026
Workflow context for AI agents with Smartsheet MCP on AWS

Smartsheet has released a generally available Model Context Protocol (MCP) Server that lets AI agents read, write, and manage Smartsheet data via a standardized protocol covering more than 40 sheet‑related tools. The server returns precisely structured, token‑efficient payloads—including row relationships, column…

Work management context for AI agents with Smartsheet MCP on AWS
AWS ML Blog· 10 min read· Oct 20, 2026
Work management context for AI agents with Smartsheet MCP on AWS

Smartsheet has launched a generally available Model Context Protocol (MCP) Server that lets AI agents read, write, and manage Smartsheet data via a standardized protocol covering more than 40 tools for sheet, row, column, and automation operations. Amazon Quick and Kiro act as MCP clients, enabling natural‑language…

Securing AI agent identity with X.509 PKI on AWS
AWS ML Blog· 17 min read· Oct 20, 2026
Securing AI agent identity with X.509 PKI on AWS

It introduces signed request envelopes, user‑signed delegation assertions, RFC 8693 audience‑scoped tokens, and real‑time OCSP revocation checking to provide cryptographic proof‑of‑possession and zero‑trust authentication for agents accessing protected tools. Engineers can adopt the provided GitHub code to embed these…

How Eon built multi-cloud investigation with an MCP Gateway on Amazon Bedrock AgentCore
AWS ML Blog· 12 min read· Oct 20, 2026
How Eon built multi-cloud investigation with an MCP Gateway on Amazon Bedrock AgentCore

Eon replaced per‑tool credentials and local Model Context Protocol (MCP) servers with a single authenticated MCP Gateway built on Amazon Bedrock AgentCore, cutting new‑engineer onboarding from roughly two days to about two minutes. The unified gateway also lets their Claude‑based autonomous SRE agent access the same…

Build a semantic layer for agentic AI on AWS with Strategy Mosaic and Amazon Bedrock AgentCore
AWS ML Blog· 28 min read· Oct 20, 2026
Build a semantic layer for agentic AI on AWS with Strategy Mosaic and Amazon Bedrock AgentCore

The article details how to construct a customer‑360 semantic layer that spans Amazon Redshift and Amazon Aurora using Strategy Mosaic. The layer is then exposed to a Strands agent running on Amazon Bedrock AgentCore via the Mosaic MCP server, enabling end‑to‑end per‑user row and column security. This integration…

Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore
AWS ML Blog· 29 min read· Oct 20, 2026
Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore

The post demonstrates how to construct a governed customer‑360 semantic layer on Amazon Redshift using AtScale, then expose that layer to a Strands agent running on Amazon Bedrock AgentCore via the AtScale MCP server. Identity propagation, row‑level security, RBAC, and full query audit are all enforced at the semantic…

Making Amazon Quick enterprise-ready: Automated, auditable cross-account resource promotion
AWS ML Blog· 12 min read· Oct 5, 2026
Making Amazon Quick enterprise-ready: Automated, auditable cross-account resource promotion

Amazon Quick’s new Model Context Protocol (MCP) server on Amazon Bedrock AgentCore automates cross‑account promotion of agents, connectors, knowledge bases, flows, and spaces by programmatically reading resources via the Amazon Quick API and upserting them into a target account. The migrator is idempotent, auditable,…

The Agent Said It Was Done. The Database Disagreed.
Hugging Face Blog· 13 min read· Oct 3, 2026
The Agent Said It Was Done. The Database Disagreed.

Microsoft’s ThinkingBox benchmark evaluates AI agents by grading the terminal database state and side‑effects they leave behind, running each of 507 stateful business workflows 20 times across 12 LLM models. In a 121,680‑trial ablation, 79,853 attempts failed the executable checks, with 67.24 % terminating cleanly yet…

Sweep thousands of leases for compliance using Amazon Quick and the Adjudicated Query pattern
AWS ML Blog· 20 min read· Oct 2, 2026
Sweep thousands of leases for compliance using Amazon Quick and the Adjudicated Query pattern

The article introduces the Adjudicated Query pattern, which couples Amazon Quick’s chat agent with a bounded Model Context Protocol (MCP) server and a deterministic rules engine to generate provably complete, defensible lease‑compliance responses. It details a reference architecture and provides a deployable AWS CDK…

CData’s AI gateway governs agents’ access to enterprise data
SiliconANGLE AI· Sep 29, 2026
CData’s AI gateway governs agents’ access to enterprise data

CData Software Inc. has launched Connect AI Gateway, an early‑access platform that extends its existing Connect AI suite to govern how AI agents select models, invoke tools, and access enterprise data. The gateway introduces a fine‑grained policy layer that can restrict model choices and tool usage per agent, ensuring…

Build a multi-account AI agent with AgentCore Gateway and MCP
AWS ML Blog· 19 min read· Sep 24, 2026
Build a multi-account AI agent with AgentCore Gateway and MCP

The article demonstrates how to architect a multi‑account AI agent system that keeps each team’s data isolated in separate AWS accounts while enabling a unified query interface. It leverages Amazon Bedrock’s AgentCore Gateway as a central orchestrator, with each line‑of‑business account exposing its data via an MCP…

From portal-hopping to instant answers: HEMA’s journey with MCP and Amazon Bedrock
AWS ML Blog· 13 min read· Sep 23, 2026
From portal-hopping to instant answers: HEMA’s journey with MCP and Amazon Bedrock

HEMA built HAL, an internal AI assistant powered by Amazon Bedrock AgentCore, that replaces a fragmented developer‑portal workflow with instant, governed answers. By leveraging the Model Context Protocol (MCP), HAL injects curated knowledge directly into the model’s context, allowing it to answer queries inside the…

Genie One MCP: Give any AI Agent the Right Business Context
Databricks Blog· 6 min read· Sep 22, 2026
Genie One MCP: Give any AI Agent the Right Business Context

Genie One MCP introduces a mechanism to supply AI agents with the appropriate business context, aiming to improve reliability of data‑driven decisions. The approach centers on a Model Context Protocol that standardizes how contextual information is attached to agent requests. This enables engineers to integrate…

Extending public sector intelligence with Agentforce and AWS
AWS ML Blog· 15 min read· Sep 22, 2026
Extending public sector intelligence with Agentforce and AWS

This article details an end‑to‑end integration that combines Amazon Bedrock Data Automation with the Model Context Protocol (MCP) to turn unstructured evidence stored in Amazon S3 into structured insights that can be queried via a Salesforce Agentforce AI agent. The solution uses S3 event notifications to trigger a…

Implementing defense-in-depth authorization for MCP tools on Amazon Quick
AWS ML Blog· 24 min read· Sep 17, 2026
Implementing defense-in-depth authorization for MCP tools on Amazon Quick

This blog post introduces a multi‑gate authorization pattern for Model Context Protocol (MCP) tool invocations on Amazon Quick, evaluating OpenID Connect (OIDC) JWT claims through four sequential gates—MFA, geographic restriction, group‑to‑role mapping, and tool‑level permission checks. It shows how to configure…

Build interactive MCP Apps using Amazon Bedrock AgentCore
AWS ML Blog· 13 min read· Sep 11, 2026
Build interactive MCP Apps using Amazon Bedrock AgentCore

Amazon Bedrock AgentCore now supports building interactive MCP Apps that can host rich HTML widgets, enabling a single server to deliver a consistent user experience across multiple AI hosts such as ChatGPT and Claude. The MCP Apps standard is host‑agnostic, so developers can write an app once and run it on any…

When One Process Becomes Too Much: Splitting a Pipeline into MCP Services
Towards Data Science· Sep 9, 2026
When One Process Becomes Too Much: Splitting a Pipeline into MCP Services

The authors demonstrate how a tightly coupled Python data‑processing pipeline can be refactored into a set of independently deployable Model Context Protocol (MCP) services, enabling each component to run in isolation while preserving the overall workflow. By exposing each stage as an MCP service, the architecture…

Connect an AgentCore Runtime hosted MCP server to Amazon Quick
AWS ML Blog· 14 min read· Aug 31, 2026
Connect an AgentCore Runtime hosted MCP server to Amazon Quick

The post demonstrates how to host an MCP server inside an AgentCore Runtime and wire it to Amazon Quick, enabling clients to share AI tools and agents without duplicating effort. By running MCP on AgentCore, the same toolset can be exposed to multiple Quick workflows, streamlining reuse across projects. The…

Agentic observability with Amazon OpenSearch Service MCP Apps
AWS ML Blog· 12 min read· Aug 25, 2026
Agentic observability with Amazon OpenSearch Service MCP Apps

Amazon OpenSearch Service now supports Model Context Protocol (MCP) Apps that extend MCP with a dual‑response pattern, returning both a concise text summary and an interactive visualization (e.g., trace waterfall, service topology) directly in the AI assistant’s chat window. This eliminates the need to switch to…

Prompt injection ranks No. 1 with OWASP and No. 12 in the incident record. The attack itself is invisible to a scan.
VentureBeat AI· 13 min read· Aug 25, 2026
Prompt injection ranks No. 1 with OWASP and No. 12 in the incident record. The attack itself is invisible to a scan.

Prompt injection has dominated the OWASP Top 10 for LLM applications as the #1 threat for three consecutive years, and a recent audit of 6,639 real‑world incidents places it at #12 on the overall incident record. Despite its high prevalence, the attack is often invisible to automated scans and is under‑represented in…

Agentic Resource Discovery (ARD): An open specification for agent discovery
AWS ML Blog· 5 min read· Aug 24, 2026
Agentic Resource Discovery (ARD): An open specification for agent discovery

AWS introduces the Agent Registry, a centralized, searchable catalog for AI agents, Model Context Protocol (MCP) servers, tools, and custom resources, and announces the Agentic Resource Discovery (ARD) open specification to enable cross‑environment discovery. The registry supports cross‑account sharing within an AWS…

EXPLORE AI NEWS

Daily hand-picked stories on LLMs, RAG, agents and production AI — curated for engineers who ship.

BROWSE NEWS

GET THE WEEKLY DIGEST

Join engineers getting the Monday signal-over-noise AI breakdown. No spam, unsubscribe anytime.

LEARN AI ENGINEERING

Curated courses, research papers, repos and tutorials built for engineers leveling up in AI.

START LEARNING