惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

MongoDB | Blog
MongoDB | Blog
大猫的无限游戏
大猫的无限游戏
H
Help Net Security
V
Visual Studio Blog
G
Google Developers Blog
Blog — PlanetScale
Blog — PlanetScale
Last Week in AI
Last Week in AI
C
Check Point Blog
D
Docker
M
MIT News - Artificial intelligence
P
Proofpoint News Feed
博客园 - 叶小钗
博客园 - 聂微东
The Cloudflare Blog
云风的 BLOG
云风的 BLOG
Martin Fowler
Martin Fowler
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
酷 壳 – CoolShell
酷 壳 – CoolShell
MyScale Blog
MyScale Blog
WordPress大学
WordPress大学
Engineering at Meta
Engineering at Meta
腾讯CDC
S
SegmentFault 最新的问题
博客园 - 【当耐特】

InfoWorld

AWS boosts CloudWatch Logs query limits by 10x to ease debugging for developers, SREs 21 LLMs tuned for special domains The new AI lock-in AWS adds Advanced Prompt Optimization tool to Bedrock Capacity markets could reshape cloud computing Four cutting-edge tools for spec-driven development Anthropic puts Claude agents on a meter across its subscriptions Notion courts developers with a platform for AI agents and workflow automation Using continuous purple teaming to protect fast-paced enterprise environments A better way to work with SQL Server Evidence-driven workflows: Rethinking enterprise process design AWS debuts Graviton-powered Redshift RG instances to cut analytics costs SAP’s AI promises last year? Most are still rolling out First look: Lemonade serves up local AI with limitations GitLab CEO sees developer tool bill increasing 100-fold Red Hat adds support for agentic AI development What’s new and exciting in JDK 26 Kill the loading spinner with local-first data and reactive SQL A networking revolution at AWS Tokenmaxxing is super dumb Hands-on with React, Supabase, and PowerSync How to add AI to an existing product (without annoying users) Your AI doesn’t need another database What happens when engineering teams reorganize around AI agents Python isn’t always easy When cloud giants meddle in markets 12 model-level deep cuts to slash AI training costs The best new features in Python 3.15 Teradata launches platform for enterprise AI agents moving beyond pilots Three skills that matter when AI handles the coding
AWS targets AI agent sprawl with new Bedrock Agent Registry
2026-04-10 · via InfoWorld

AWS is expanding its Amazon Bedrock AgentCore portfolio with a new managed service, Agent Registry, designed to help enterprises catalog, manage, and govern their growing fleets of AI agents and their associated tools.

The service provides a unified directory of agents, capturing metadata such as capabilities, identities, and integrations, and is designed to support agents built with different models and frameworks, AWS wrote in the offering’s documentation.

The move addresses a growing structural gap around AI agent sprawl in enterprises that has emerged as they try to move pilot cases to production, analysts say.

“Agent Sprawl is an emerging structural problem. What we see consistently across enterprises is that agents proliferate much faster than traditional applications because they are easier to build. As a result, ownership becomes ambiguous as they move into production while increasing costs, risks, and duplication until finance, security, or related incidents force attention,” said Gaurav Dewan, research director at Avasant.

With the addition of a centralized Agent Registry-like offering, discovery, orchestration, governance, lifecycle management, and standardization of agents become easier, Dewan added, noting that in effect, a registry transforms agents from isolated artifacts into managed, composable enterprise assets.

A control plane play with trade-offs?

The service comes with what analysts describe as a “strategic” limitation: While it can track agents interacting with external systems, the registry itself operates within AWS.  

This approach, according to Forrester principal analyst Charlie Dai, reinforces the company’s push to position Bedrock as the control plane for enterprise AI agent deployment and oversight.

More so because other cloud providers are taking a similar approach. 

Google Cloud, for instance, is extending Vertex AI with capabilities to orchestrate and monitor agents, including a governance layer within Vertex AI Agent Builder and integrations with its Agent Registry via Apigee.

Microsoft, meanwhile, is positioning Azure AI and Copilot Studio as a unified platform for building and governing enterprise AI agents, complemented by Agent 365 and Entra Agent ID for discovery and identity management.

In fact, Dewan cautioned enterprise teams planning to embrace AWS’ Agent Registry because of its close integration with AWS-native services, particularly in areas such as identity and runtime.

“As a result, while the service will natively index and manage agents deployed within AWS environments, integration with external or on-prem agents will likely require manual registration. Cross-cloud or federated discovery capabilities are not yet clearly established,” Dewan said.

This limitation, in itself, could introduce a new risk: registry sprawl across hyperscalers, Dewan noted, adding that enterprises adopting AWS, Microsoft, and Google registries in parallel could end up recreating the very fragmentation these tools are meant to solve.

Accessing the Agent Registry and adding agents

The registry is accessible through multiple entry points, including the AgentCore console, APIs, SDKs, and even as a Model Context Protocol (MCP) server, allowing compatible clients and developer tools to query it directly.

This multi-access design, according to an AWS blog post, is deliberate as it allows teams to integrate the registry into existing development environments or build custom discovery interfaces using OAuth-based authentication, without being tied strictly to AWS-native tooling.

When it comes to adding agents and related resources, AWS provides two primary approaches.

The first is a manual registration path, where developers or platform teams create records via the console, APIs, or SDKs by supplying structured metadata such as capabilities, ownership, compliance attributes, and usage documentation.

The second is a more automated ingestion route, where teams can point the registry to an MCP or Agent2Agent (A2A) endpoint, allowing it to pull in agent details automatically, AWS wrote in the blog post.

Agent Registry is currently available in preview across five AWS Regions, including US West (Oregon), Asia Pacific (Tokyo), Asia Pacific (Sydney), Europe (Ireland), and US East (N. Virginia).

The service is expected to start supporting external registries soon, AWS said.

That, the hyperscaler added, will help enterprises connect multiple registries and search across them as one.

“You will be able to define categories and taxonomies that match how your organization thinks about agents, backed by structured metadata schemas capturing ownership, compliance status, cost center, and whatever else your governance model requires,” it further detailed in the blog post.