โ๏ธAWS Machine Learning BlogโขStalecollected in 26m
Extending MCP support for Amazon Bedrock AgentCore Gateway
๐กEnterprise-grade security and governance for MCP deployments on AWS infrastructure.
โก 30-Second TL;DR
What Changed
Centralized credential management for MCP servers at scale
Why It Matters
This update simplifies the operational overhead of managing agentic AI architectures in enterprise settings. It allows teams to scale MCP deployments securely without compromising on governance or visibility.
What To Do Next
Review your current MCP server architecture and evaluate if Amazon Bedrock AgentCore Gateway can replace your custom middleware for centralized security and logging.
Who should care:Enterprise & Security Teams
Key Points
- โขCentralized credential management for MCP servers at scale
- โขEnhanced observability into tool usage across enterprise teams
- โขSecurity features to prevent data exfiltration in production environments
- โขFine-grained access control for MCP-based agentic workflows
๐ง Deep Insight
Web-grounded analysis with 24 cited sources.
๐ Enhanced Key Takeaways
- โขThe Model Context Protocol (MCP) is an open standard, inspired by the Language Server Protocol, that enables LLM applications to securely connect with external data sources and tools, streamlining integrations by reducing the need for N*M custom connections to N+M standardized ones.
- โขAmazon Bedrock AgentCore Gateway now supports server-side tool execution through its integration with the Bedrock Responses API, allowing models to discover, select, and execute tools directly during inference without requiring client-side orchestration, thereby reducing application complexity and latency.
- โขAgentCore Identity, a component of AgentCore, provides centralized agent identity management, including a secure token vault for OAuth 2.0 tokens and API keys, and supports various OAuth 2.0 flows (2LO and 3LO) for secure, identity-aware authorization when agents access external services.
- โขAgentCore Observability offers deep visibility into agent performance by emitting telemetry data in OpenTelemetry (OTEL)-compatible format, enabling detailed visualizations of agent workflow steps, including token usage, latency, session duration, and error rates, through Amazon CloudWatch dashboards and integrations with third-party tools like Langfuse and Dynatrace.
- โขAgentCore Gateway can unify multiple existing MCP servers, whether custom-built, publicly available, or open-source, under a single control plane, facilitating consistent tool discovery, access, and management across different organizational units or product features.
๐ ๏ธ Technical Deep Dive
- Model Context Protocol (MCP): An open standard built on JSON-RPC 2.0 foundations, designed to standardize how LLM applications interact with external tools, environments, and memory. It defines how AI agents discover capabilities, subscribe to data streams, and maintain contextual state across interactions. MCP utilizes OAuth for authentication and incorporates a permission model where the host controls client access to specific servers and tools, ensuring data isolation between sessions.
- AgentCore Gateway: This service acts as a centralized hub, converting various input types such as OpenAPI specifications, Smithy models, and AWS Lambda functions into MCP-compatible tools. It provides comprehensive ingress and egress authentication, supporting the OAuth 2.0 Authorization Code flow via AgentCore Identity. The Gateway integrates with the Bedrock Responses API to enable server-side tool execution and connects to external MCP servers using the
SynchronizeGatewayTargetsAPI for protocol handshakes and capability indexing. - AgentCore Identity: A dedicated service for managing AI agent identities. It features a secure token vault for encrypting and storing OAuth 2.0 Access and Refresh tokens, API keys, and OAuth client secrets using AWS Key Management Service (KMS). AgentCore Identity supports both OAuth 2.0 client credentials grant (2LO) and authorization code grant (3LO), facilitating identity-aware authorization by securely passing user context to agent code.
- AgentCore Observability: This component is designed for tracing, debugging, and monitoring agent performance in production. It emits telemetry data in an OpenTelemetry (OTEL)-compatible format, allowing for integration with existing monitoring stacks. Detailed visualizations of agent workflow steps, including metrics like token usage, response latency, session duration, and error rates, are provided through Amazon CloudWatch dashboards. It also supports integration with third-party observability platforms like Langfuse and Dynatrace.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
Standardized protocols like MCP will become essential for widespread AI agent adoption.
The open nature of MCP and its adoption by major cloud providers like AWS will drive industry-wide standardization, reducing integration friction and fostering a more interoperable ecosystem for AI agents.
Enterprise AI agent deployments will increasingly prioritize robust security and governance features.
The focus on centralized credential management, fine-grained access control, and comprehensive observability in AgentCore Gateway addresses critical enterprise requirements, accelerating the deployment of AI agents in sensitive and regulated environments.
The complexity of AI agent workflows will grow, necessitating advanced orchestration and multi-agent collaboration capabilities.
With foundational security and integration challenges being addressed, enterprises will shift towards building more sophisticated AI systems involving multiple specialized agents working in concert to automate complex business processes.
โณ Timeline
2023-04-13
Amazon Bedrock announced in preview.
2023-09-28
Amazon Bedrock General Availability.
2023-11-28
Agents for Amazon Bedrock General Availability.
2024-11-05
Model Context Protocol (MCP) initial stable version released.
2025-10-13
Amazon Bedrock AgentCore General Availability, including Gateway, Identity, and Observability services.
2026-02-24
AgentCore Gateway enables server-side tool execution with Amazon Bedrock Responses API.
๐ Sources (24)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: AWS Machine Learning Blog โ