Tech Meridian ← ENTITY INDEX
PROMY MERIDIAN RU

COMPANY · ENTITY #437

Model Context Protocol (MCP)

Related event timeline, sources and context from the news index.

EVENT TIMELINE

45

COMPANIES · 1 SOURCE · AWS Machine Learning

MoEngage uses Amazon Bedrock AgentCore Gateway to expose 400 private APIs as agent tools

MoEngage converted roughly 400 internal REST APIs across seven AWS Regions into permission-scoped agent tools using AgentCore Gateway (part of Amazon Bedrock AgentCore). The approach centralized MCP handling under MoEngage’s AI Platform team, avoided any public endpoints, preserved identity injection server-side, and cut estimated platform engineering effort by about six months while reducing API onboarding from days to hours.

6.0

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon publishes Enterprise Agentic AI Platform blueprint built on Bedrock AgentCore

Amazon published a reference architecture and a companion "Enterprise Agentic AI Platform Blueprint" built on Amazon Bedrock AgentCore that describes a multi-account, governed platform pattern with per-workstream deployments, separate inference and tool gateways, identity and cost controls, and an evaluation pipeline; the blueprint is available as deployable infrastructure-as-code.

7.0

COMPANIES · 2 SOURCES · Google · TechCrunch AI

Google launches Connected Apps Platform preview for Gemini and Search

Google announced the preview of a Connected Apps Platform that gives developers a single place to build, register, test and manage apps that integrate with the Gemini app and Google Search. Developers integrate via a cloud-hosted MCP server or Android MCP using AppFunctions APIs and use the Google Connected Console to submit, control branding, and monitor traffic; access will open in phases and partners such as Adobe, Calm, Peloton, Experian and Localiza are already live in Gemini with some integrations coming to Search.

7.0

CODING · 1 SOURCE · AWS Machine Learning

Amazon Bedrock documents prompt-caching patterns for running OpenAI models in agents

An Amazon Bedrock post explains implicit and explicit prompt caching for agent workloads running OpenAI models, showing how to save the model’s prefill (KV) state for reusable prefixes (e.g., tools and instructions) using prompt_cache_breakpoint and prompt_cache_key to cut token costs and latency; it notes caching requires exact identical text and a reusable prefix of at least 1,024 tokens. The post also discusses combining implicit caching (Bedrock-chosen breakpoints) with explicit markers to share starting state across sessions and subagents.

5.0

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS documents how to deploy Amazon Quick in HIPAA-eligible environments

AWS published guidance explaining that Amazon Quick is a HIPAA-eligible service and describing how healthcare organizations can use its agentic chat agents, connectors, and Model Context Protocol (MCP) to process PHI within HIPAA-eligible AWS accounts. The post details available controls (ACL-aware retrieval, chat/feedback logging, action controls), shared-responsibility limits of the AWS BAA, and best practices for governance and irreversible configuration choices.

7.0

COMPANIES · 1 SOURCE · AWS Machine Learning

Rescale builds automated test-triage agent using Strands Agents and AI-DLC

Rescale used the open-source AI-Driven Development Lifecycle (AI-DLC) and Strands Agents to build an agent that classifies and routes nightly end-to-end test failures. The agent runs in Argo Workflows on Amazon EKS, invokes Claude Opus 4.8 via Amazon Bedrock, queries logs through the Model Context Protocol (MCP), and produced a staged deployment in five days and production deployment within weeks.

6.0

COMPANIES · 1 SOURCE · AWS Machine Learning

Smartsheet launches GA MCP Server on AWS to let AI agents read and write sheets

Smartsheet has launched a generally available (GA) Model Context Protocol (MCP) Server hosted on AWS that lets AI assistants read, write, and manage Smartsheet data via a standardized MCP interface. The server supports Streamable HTTP transport, MCP spec versions 2025-06-18 and 2025-03-26, three regional endpoints, OAuth 2.0 or Personal Access Token auth, and is shown integrated with Amazon Quick and the Kiro IDE as MCP clients.

6.0

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS blog demonstrates conversational supply-chain workflows using Bedrock AgentCore and Amazon Quick

An AWS blog post walks through a pattern using Amazon Bedrock AgentCore (the agentic intelligence layer) together with Amazon Quick (a conversational interface) to query live systems of record via Model Context Protocol (MCP) tools and orchestrate end-to-end supply‑chain tasks like disruption impact assessment and emergency RFQs. The post describes architecture, MCP connectors, and a step‑by‑step conversational demo that returns grounded, approval‑gated actions from multiple backends.

6.0

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS Agent Registry enables runtime discovery of MCP servers and agent skills

AWS describes AWS Agent Registry, a managed catalog for registering and semantically searching Model Context Protocol (MCP) servers, agents, and agent skills so orchestrators can discover and instantiate tools at runtime rather than hardcoding endpoints. The post includes a Python reference that wires Amazon Bedrock AgentCore Gateway, AgentCore runtime A2A agents, and an orchestrator that searches the registry, constructs clients from descriptors, and invokes capabilities (authentication must be obtained separately).

6.0

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS and Keyfactor publish X.509 PKI reference architecture for AI agents

AWS and Keyfactor released a joint reference architecture showing how to use Keyfactor EJBCA with AWS services (including Amazon Bedrock and the AWS Strands Agents SDK) to provide cryptographic X.509 identities for AI agents. The design uses signed request envelopes, user-signed delegation assertions, RFC 8693 audience-scoped tokens, and real-time OCSP revocation; example source code is available on GitHub.

7.0

COMPANIES · 1 SOURCE · TechCrunch AI

Figma adds 20 MCP partners while co‑founder says MCP is 'not a perfect protocol'

At its Config conference in India, Figma announced support for 20 new Model Context Protocol (MCP) partners — including Notion, Mistral, Meta Muse, Junie by JetBrains, ClickUp and others — even as co‑founder Dylan Field said MCP is "not a perfect protocol" and flagged concerns about asymmetric context sharing; Figma also said it will support Pi and cited risks from exposing more user data via MCP.

6.0

RESEARCH · 1 SOURCE · arXiv cs.AI

Structured Authority Injection (SAI): protocol for persistent human overrides in LLM agents

The paper proposes Structured Authority Injection (SAI), a protocol that routes operator corrections through typed, schema-validated Model Context Protocol (MCP) tool calls and maintains a ranked authority context prepended to each LLM prompt. SAI defines three override classes (Constraint, Priority, Strategy) with distinct persistence and precedence enforced via a directed authority channel, requires no model fine-tuning, and evaluation across nine LLMs and eight scenarios shows reduced ambiguous directive interpretation and mitigated authority decay (with prompt overhead reductions up to ~38% in the strongest models), though benefits vary by model and scenario.

6.0

RESEARCH · 1 SOURCE · arXiv cs.AI

Which Metadata Matters for LLM Tool Selection in MCP (arXiv:2610.14334v1)

This arXiv paper analyzes metadata for 367,174 MCP tools and runs controlled, single-factor metadata interventions across five LLMs and five tasks, estimating position-adjusted effects with conditional logistic regression. It finds parameter-schema differences drive the largest and most consistent shifts in which tool an LLM selects, descriptions that explain parameters increase selection, names have weaker effects, and explicit request-compatibility can override position biases.

7.0

RESEARCH · 1 SOURCE · NVIDIA Developer

Siemens Fuse EDA agent reliability raised from ~55% to ~95% using NVIDIA NeMo Gym

Siemens used NVIDIA NeMo Gym to evaluate and improve its Fuse EDA AI Agent for standard-cell characterization, boosting a composite verification score from about 55% to ~95% across a representative benchmark. Refining agent skills and Model Context Protocol (MCP) tool descriptions based on verifier diagnostics also cut agent turns from 155 to 53 and reduced output tokens from ~81K to ~29K.

7.0

COMPANIES · 1 SOURCE · AWS Machine Learning

Anchor Browser integrates with Amazon Bedrock AgentCore payments to handle x402 machine-to-machine payments

Anchor Browser now integrates with Amazon Bedrock AgentCore payments so AI agents can autonomously respond to HTTP 402 Payment Required challenges using the x402 protocol. The integration, developed with the AWS Generative AI Innovation Center Partner Agent Factory and available on AWS Marketplace, manages wallets, spending limits, and CloudWatch observability so agent workflows continue without manual intervention.

6.0

COMPANIES · 1 SOURCE · TechCrunch AI

Google launches agentic Gemini unified agent for enterprise

At a Google Cloud event Google announced a unified, agentic version of Gemini aimed first at businesses via Gemini Enterprise; the agent can accept objectives, plan and execute tasks, use custom skills and tools, and connect to internal systems. It offers a model picker (including Anthropic’s Claude), per-agent Workspace accounts, a tasks inbox and audit trails, MCP compatibility, and integrations with systems like Google Workspace, Microsoft 365, Slack, BigQuery and Snowflake.

8.0

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS publishes tutorial to build a voice travel concierge with Bedrock AgentCore and Nova 2.5 Sonic

AWS published a how-to demonstrating a voice travel concierge built with Amazon Bedrock AgentCore (runtime and Gateway), Amazon Nova 2.5 Sonic for real‑time speech, and Amazon Bedrock Knowledge Bases, using the AWS CDK and the Model Context Protocol (MCP) to connect the agent to backend services. The post details architecture, components (authentication, data stores, Lambdas, API Gateway), and a sample airline implementation that keeps the agent loosely coupled to existing systems.

5.0

RESEARCH · 1 SOURCE · Ars Technica

Researcher finds 'protocol pivoting' vulnerabilities in MCP that enable agent-to-agent exploits

Independent researcher Syed Anas Mohiuddin demonstrated a class of attacks he calls “protocol pivoting” that exploits trust gaps in the Model Context Protocol (MCP) to pass malicious instructions between AI agents; he tested agents from Google, Rapid7, JP Morgan Chase, Weviate, the French interministerial digital directorate and US federal systems. Mohiuddin’s proofs-of-concept include server-side request forgery (SSRF) and prompt-injection chains; Rapid7 patched CVE-2026-97228 (severity 2.7) and Google fixed an issue in googleapis/mcp-toolbox (rated higher) by adding IP allow-lists and stricter URL validation.

8.0

CODING · 1 SOURCE · AWS Machine Learning

Amazon SageMaker launches aws-ai-ml agent skill for inference optimization

Amazon introduced the aws-ai-ml skill available through the Agent Toolkit for AWS, which plugs into MCP‑compatible coding agents (e.g., Kiro, Claude Code, Codex) to benchmark SageMaker endpoints, recommend deployment configurations, compare performance runs, and generate executable SageMaker Python SDK v3 code. The skill can be used locally via the Agent Toolkit or inside Amazon SageMaker Studio JupyterLab and requires AWS credentials with SageMaker API permissions.

6.0

CODING · 1 SOURCE · AWS Machine Learning

Amazon publishes Quick Resource Migrator — MCP server on Bedrock AgentCore to automate cross-account promotion

Amazon posted a sample Quick Resource Migrator — an idempotent, auditable Model Context Protocol (MCP) server running on Amazon Bedrock AgentCore that automates promotion of Amazon Quick resources (agents, connectors, knowledge bases, flows, spaces) across AWS accounts. The tool reads resources and permissions via the Amazon Quick API, upserts them into a target account, writes versioned backups to S3, and provides a read-only preview; source code is available in the aws-samples repository.

6.0

CODING · 1 SOURCE · AWS Machine Learning

Amazon Bedrock AgentCore Gateway adds secure Web Search access to Claude Desktop

AWS shows a step-by-step walkthrough to connect Claude Desktop on Amazon Bedrock to AgentCore Gateway’s managed Web Search, using JWT-based inbound authentication with Amazon Cognito and AWS IAM Identity Center so query traffic stays within AWS. The guide covers configuration, SAML/OAuth federation, and notes regional availability in us-east-1, eu-west-1, and ap-northeast-1.

6.0

COMPANIES · 1 SOURCE · NVIDIA Developer

NVIDIA releases VSS Blueprint 3.3 with Build Vision Agent skill and Adaptive EVS

NVIDIA’s Metropolis Video Search and Summarization (VSS) Blueprint 3.3 adds the vss-build-vision-ai skill to compose multi-workflow visual AI deployments from a single prompt and introduces Adaptive Efficient Video Sampling (EVS) to prune redundant visual tokens and batch VLM work. NVIDIA cites a demo that deployed a bottling-line agent in under 30 minutes on an RTX PRO 6000 Blackwell and reports up to ~80% fewer VLM input tokens for a 60-minute summary and a 46% increase in concurrent real-time VLM streams in their tests; the repository and agent skills are available and a live demo is scheduled for Oct. 1 at 9 a.m. PT.

7.0

RESEARCH · 1 SOURCE · Hugging Face

ProvenanceGuard: source-aware factuality verification for MCP-based LLM agents

The paper ProvenanceGuard proposes a post-generation verifier that preserves source identity from Model Context Protocol (MCP) tool traces to check both whether each claim is supported and whether it is attributed to the correct tool output. The method decomposes answers into claims, finds the most relevant MCP source for each, verifies support (using models in the authors' local setup such as MiniLM and DeBERTa), compares claimed attribution, and can block or prompt source-grounded repairs; experiments used 281 real medical-agent traces.

7.0

COMPANIES · 1 SOURCE · TechCrunch AI

Shopify opens checkout to browser-based AI agents

Shopify announced WebMCP support for checkout (including Shop Pay), allowing browser-based AI agents to inspect, update, and complete purchases on eligible merchants’ storefronts via three new tools: get_checkout, update_checkout, and complete_checkout. The change extends existing WebMCP storefront/cart capabilities and leverages Shopify’s Universal Commerce Protocol (UCP); Shopify also offers a hosted MCP server and partners such as Muse and Instinct for agentic commerce.

7.0

RESEARCH · 1 SOURCE · arXiv cs.AI

Model Context Protocol and Eunomia Agent propose mediation layer to connect LLM agents with Data Spaces (arXiv:2609.30341v1)

The paper proposes an architectural mediation approach using the Model Context Protocol (MCP), implemented in the Eunomia Agent, to translate data space capabilities into schema-driven tools LLM agents can discover and invoke. A prototype demonstrates end-to-end interactions for catalog discovery, metadata retrieval, and data service invocation without modifying existing data space components, aiming to preserve governance and interoperability.

6.0

COMPANIES · 1 SOURCE · InfoQ AI, ML & Data Engineering

AWS: MCP specification makes protocol stateless and drops session affinity

The updated Model Context Protocol (MCP) specification removes the initialize/initialized handshake and the Mcp-Session-Id header, allowing requests to be routed independently to any server instance and eliminating the protocol-level requirement for sticky sessions or shared session stores. AWS maps the change to its Well-Architected Agentic AI guidance, highlights serverless and conventional load-balancer deployment patterns, notes new headers and tracing/caching controls (Mcp-Method, Mcp-Name, W3C Trace Context, ttlMs, cacheScope), and warns that clients must handle retries and idempotency because stream resumability was removed.

7.0

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon Bedrock: multi-account AI agent architecture using AgentCore Gateway and MCP

Amazon’s post describes a reference architecture for building multi-account AI agents that query data in-place across multiple AWS line-of-business (LOB) accounts without centralizing datasets. It shows how a central platform account can host agents and Bedrock inference while AgentCore Gateway (with MCP, AgentCore Identity, Policy in AgentCore/Cedar, and optional Okta auth) aggregates LOB MCP servers, provides unified tool discovery, authorization, observability, and applies Bedrock guardrails and RAG-backed knowledge retrieval.

6.0

COMPANIES · 1 SOURCE · AWS Machine Learning

HEMA builds internal assistant HAL using Model Context Protocol and Amazon Bedrock AgentCore

Dutch retailer HEMA created HAL, an internal AI assistant that consolidates fragmented internal knowledge and delivers answers inside existing tools by exposing sources as MCP tools and running agents on Amazon Bedrock AgentCore. Identity is anchored with Microsoft Entra ID, and AgentCore features (Gateway, Runtime, Memory, Guardrails) were used to avoid custom MCP server infrastructure.

5.0

COMPANIES · 1 SOURCE · InfoQ AI, ML & Data Engineering

Morgan Stanley uses CALM Architecture-as-Code to integrate MCP and A2A into its API program

At QCon London, Morgan Stanley engineers Jim Gough and Andreea Niculcea demonstrated how the firm has modernized its API program using an Architecture-as-Code approach called CALM, integrating the Model Context Protocol (MCP) and Agent-to-Agent (A2A) communications. They showed demos of MCP-enabled flows, automated governance via deployment gates, and zero-downtime platform upgrades to scale agentic workflows and production APIs (over 110 APIs deployed to date).

5.0

CODING · 1 SOURCE · AWS Machine Learning

Strands Evals SDK and Amazon Bedrock AgentCore add skill-focused evaluators

Strands Evals SDK and Amazon Bedrock’s AgentCore Evaluations now include skill-focused evaluators that measure whether an agent chose the appropriate skill (Skill Selection Accuracy) and how fully it followed that skill’s instructions (Skill Instruction Following). Strands also provides a deterministic SkillInvoked check; evaluations can run from Strands trajectories or OpenTelemetry traces and be used via the AgentCore CLI.

6.0