MCP was supposed to solve the agent tooling problem. It missed a step.
Connecting an AI agent to a tool is relatively straightforward. Things get more complicated once an organization has hundreds or The post MCP was supposed to solve the agent tooling problem. It missed a step. appeared first on The New Stack .
The Model Context Protocol (MCP) was intended to streamline the process of connecting AI agents with external tools and data. However, it falls short by assuming clients already know which server they need to use, which is problematic when organizations have resources scattered across various clouds, platforms, and SaaS systems.
AWS highlighted the open specification for Agentic Resource Discovery (ARD) in its August 31 Weekly Roundup, describing it as "DNS, but for agents." While ARD allows agents to search across different registries for required resources, it is not an AWS technology but rather a collaborative effort involving engineers from companies like Google, Microsoft, Hugging Face, Cisco, Databricks, GitHub, GoDaddy, Nvidia, Salesforce, ServiceNow, and Snowflake.
Unlike ARD, MCP does not address the discovery step, instead expecting clients to know the server they want ahead of time. ARD promotes a governance model where companies can maintain their catalogs and policies while routing searches to other ARD-compatible services, allowing enterprises to keep internal resources private while still accessing approved external catalogs when necessary.
AWS refers to this approach as "describe once, discover everywhere." Currently, the v0.91 proposal uses JSON-LD and a REST interface, with a POST /search endpoint to search by task. The Route 53 engineers involved in ARD development have shaped the specification, while AWS's Agent Registry could benefit from ARD by incorporating resources running elsewhere without requiring companies to register everything with AWS.
However, the governance of ARD remains unclear, with discussions ongoing about board terms and membership, and potential neutral organization transitions such as the W3C or an AI foundation.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
