All news with #bedrock agents tag

🔎 Web Search on Amazon Bedrock AgentCore is now available in four additional AWS Regions: US East (Ohio), US West (Oregon), Asia Pacific (Sydney), and AWS GovCloud (US-West). This fully managed tool provides AI agents with current web grounding to improve research, summarization, and question-answering. The expansion reduces latency by allowing retrieval closer to applications and end users. Government customers can now use Web Search in AWS GovCloud (US-West) to address compliance and data residency needs.

🧾 Amazon Bedrock AgentCore now charges for short-term memory (STM) based on consumption rather than a flat per-event fee. STM is billed across three dimensions: ingestion ($1.00/GB), retrieval ($0.20/GB), and storage ($0.10/GB-month), with per-event billing subject to a 12 KB minimum and 64 KB maximum billable size. Storage costs reflect actual event size and can be managed via TTL settings; long-term memory pricing remains unchanged. The model is available in all Regions where AgentCore Memory is offered.

🤖 Developed jointly by AWS and OpenAI, Bedrock Managed Agents (BMA) is an AWS-native implementation of OpenAI's Agents API that runs entirely inside AWS. BMA maintains state through durable sessions, manages tool selection and execution, and allows reusable skills and MCP-connected tools. Each agent uses its own IAM role, supports human approvals for consequential actions, and logs supported API activity to AWS CloudTrail. The preview is available in select US regions with no additional BMA charge beyond consumed AWS resources.

🔒 Two vulnerabilities in Amazon Bedrock AgentCore's Python SDK could let attackers execute commands inside Code Interpreter sandboxes and access AWS credentials. BeyondTrust detailed that crafted package names and pip extras syntax could bypass validation, tracked as CVE-2026-12530 and CVE-2026-16796. AWS patched the issues in versions 1.6.1 and 1.18.1, and urged upgrades and stricter handling of untrusted package names.

🚀 The new AgentCore Runtime for Amazon Bedrock AgentCore is now generally available, introducing elastic memory management that reclaims unused memory and consistent cold start times regardless of container image size. It preserves serverless benefits—no pre-provisioning, scale-to-zero, hardware-enforced isolation, and pay-for-use—while lowering costs and speeding startup. Sessions begin with a compact memory profile and expand on demand; idle memory is reclaimed. Snapshot-based cold start restoration delivers P75 cold starts of ~1.9–2.0s across 200 MB–2 GB images in tests. The runtime is available in select regions and can be enabled by setting platformVersion to V2 when creating or updating a runtime.

🔔 Amazon Bedrock AgentCore Memory now offers the IngestData API to submit content directly for long-term memory extraction without first creating a short-term memory event. The API accepts conversational and JSON payloads along with optional metadata, fans content out to configured long-term strategies, and makes resulting records available via existing retrieval operations. Developers can verify and manage extraction results with ListMemoryRecords, RetrieveMemoryRecords, Kinesis streaming for notifications, and ListMemoryExtractionJobs for redrives.

🔒 Amazon Bedrock AgentCore Identity now includes a managed consent portal that removes the need for custom OAuth callback infrastructure when connecting agents to third-party services. The portal is provided per AgentCore Gateway with a hosted web client and credential provider list, enabling platform admins to share a consent URL and users to self-service view connection status. The feature supports agent IDE-based clients that cannot natively handle OAuth consent flows and is available in all commercial regions where Bedrock AgentCore Identity is offered.

🔗 Amazon Quick now integrates with the AWS Agent Registry to let users discover, enable, and share MCP servers and agents directly within their Amazon Quick workspace. Connection details are pre-populated from the registry, enabling quick enablement and distribution for chat, agents, apps, flows, and research. The integration is available in all regions where Amazon Quick and Amazon Bedrock AgentCore are supported, and administrators can connect via Manage account > Permissions > AWS Agent Registry.

🔎 AWS Agent Registry is now generally available, offering a private, governed catalog and discovery layer for agents, skills, tools, MCP servers, and custom resources across organizations. It provides semantic and keyword search, a browse experience, and access via the console, AWS CLI, and AWS SDK. Registry integrates with Amazon Bedrock AgentCore, Amazon Quick, and Kiro IDE, and exposes an MCP server interface for IDE queries. New enterprise features include infrastructure-as-code support, tagging, AWS RAM sharing, auto-discovery of AgentCore runtimes, and a Quick Connectors integration.

🔒 Amazon Bedrock’s AgentCore Memory now supports fine-grained access control (FGAC), allowing per-user and per-tenant memory isolation via AgentCore Gateway without custom authorization code. Administrators can configure OAuth (JWT) authentication and attach Cedar policies to restrict access by caller identity, namespace claims, or specific Memory operations. The feature moves access enforcement into the infrastructure using cryptographic identity proof, and is implemented through the managed AgentCore Memory connector exposing 12 Memory operations as Cedar actions.

🧭 Amazon Bedrock AgentCore Memory now supports flexible namespace variables that let developers scope long-term memories along custom application dimensions such as organization, tenant, team, or environment. Define up to five keys on a memory resource, reference them in a strategy's namespace template, and supply values at runtime via the CreateEvent API so the service substitutes them during memory extraction. The feature helps multi-tenant and complex-hierarchy applications avoid duplicate strategies and provides fine-grained control over memory organization and access. Flexible namespace variables are available today in all Regions where AgentCore Memory is generally available, at no additional cost.

🚀 Amazon Bedrock AgentCore is now available in US West (N. California) and Asia Pacific (Hyderabad), enabling developers to build and run agents closer to end users. The platform supports agent runtime, identity and access control, policy management, session persistence, tool connectivity, evaluations, and observability at launch. AgentCore emphasizes secure infrastructure-level enforcement so agents cannot bypass controls. Regional availability reduces latency and improves performance for local customers.

🧭 Amazon Bedrock AgentCore Memory now accepts non-conversational JSON payloads on the CreateEvent API, enabling extraction of long-term memories from structured data. Developers can submit JSON events up to 100 KB—such as behavioral events, activity logs, and system events—without converting them into synthetic conversations. The pipeline applies all four extraction strategies: semantic, user preference, summarization, and episodic, and the feature is available now in all supported regions.

🔎 Amazon Bedrock's AgentCore Web Search now supports per-request domain filtering and published-date filtering, letting agents control which sites and time windows they search. Runtime domain filtering enables include/exclude lists for trusted or blocked sources without admin reconfiguration, while published-date filtering constrains results to specific inclusive date ranges. Admins gain gateway-level allowlists and an increased domain cap of up to 100 domains per list, and the Web Search Tool is expanding availability to eu-west-1 and ap-northeast-1.

🛡️ This post demonstrates patterns for propagating user authorization context when building AI agents with Amazon Bedrock AgentCore, ensuring each user only sees data they’re allowed to access. It explains how to embed department or custom claims in tokens via Amazon Cognito pre token generation triggers and how the AgentCore Runtime inbound JWT authorizer validates those claims before invoking agent code. The guidance shows moving enforcement into infrastructure—using STS session tags, per-request AssumeRoleWithWebIdentity for DynamoDB, metadata filters for Bedrock Knowledge Bases, and on‑behalf‑of token exchange for external SaaS—to reduce risk from compromised agents.

🔒 This post explains how to implement a request Lambda interceptor in Amazon Bedrock AgentCore Gateway to support legacy HTTP Basic Authentication for downstream tool APIs. It shows retrieving a service account credential from AWS Secrets Manager, validating inbound JWTs, and constructing a Basic Auth header while keeping credentials isolated from the agent. The article emphasizes that Basic Auth is outdated and recommends modern authentication like OAuth 2.0 or OpenID Connect, presenting this pattern only as an interim integration.

🔔 Amazon Web Services announces general availability of AgentCore payments, part of Amazon Bedrock AgentCore, enabling AI agents to discover, access, and pay for paid APIs, MCPs, and content with minimal code. The capability integrates with Coinbase and Stripe Privy wallets, supports the Machine Payment Protocol (MPP), and offers payment orchestration, configurable infrastructure-level limits, and end-to-end observability. GA features include Quick Create for Coinbase credentials, a curated Coinbase Bazar MCP server of x402 pay-per-use endpoints, and support for the x402 "upto" scheme for dynamic pricing.

🔧 The Amazon Cognito (aws-auth) skill is now included in the Agent Toolkit for AWS, enabling AI coding agents to set up, configure, secure, and troubleshoot Amazon Cognito using best-practice workflows. The skill supports user pools, app clients, OAuth 2.0 flows, token and JWT authorizer management, passkey/WebAuthn enrollment, threat protection, Lambda triggers, and identity pools. When used with the AWS MCP Server, commands run with IAM guardrails and CloudTrail audit logging; it also works standalone via the AWS CLI.

🛡️ Amazon Bedrock AgentCore is now available in AWS GovCloud (US-West), enabling regulated organizations to build context-aware agents with controls for production scale. AgentCore memory provides short-term conversational context and long-term persistent insights without complex infrastructure. Policy features let teams author natural-language policies that convert to Cedar and enforce tool access at an AgentCore gateway. The managed harness simplifies deployment by declaring models, tools, and instructions via configuration and running agents with a few API calls.

🚀 Amazon Web Services (AWS) announced general availability of runtime instances for Amazon Bedrock AgentCore, enabling agents to run on customer-managed EC2 instances without manual infrastructure management. The feature complements existing microVM-based runtimes and supports diverse EC2 families including GPU-accelerated, memory-optimized, and compute-optimized types. Runtime instances allow long-running agent sessions up to 14 days and integrate with the AWS Console, CLI, SDKs, and APIs to create capacity providers and attach agents. Availability spans multiple regions and billing includes EC2 costs plus management fees, with documentation and pricing details provided by AWS.