#AmazonBedrock

www.synestesia.uk

Read the full article →

aws.amazon.com

Amazon Bedrock AgentCore adds memory, policy, and harness in AWS GovCloud (US-West)

Today, Amazon Bedrock AgentCore introduces new capabilities in AWS GovCloud (US-West). AgentCore is the platform to build, connect, and optimize agents. With these capabilities, teams operating in regulated environments can build context-aware agents and take them from prototype to production faster, with the controls needed to scale across their organization.  AgentCore memory gives agents short-term memory for immediate conversation context and long-term memory that extracts persistent insights and preferences across sessions. Agents deliver more intelligent, personalized experiences without teams needing to manage complex memory infrastructure. Policy in AgentCore provides centralized, fine-grained controls for agent-tool interactions that operate outside agent code. Teams author policies in natural language that automatically convert to Cedar, the AWS open-source policy language. Policies attach them to an AgentCore gateway that evaluates each request before allowing or denying tool access.  The managed harness lets developers declare an agent's model, tools, and instructions through configuration and run it in just a few API calls without writing orchestration code, while AgentCore handles the environment, compute, memory, identity, and observability.    To learn more about AgentCore capabilities, visit the https://aws.amazon.com/bedrock/agentcore/?trk=f7596381-283b-40e2-90fe-22e22b18daf5&sc_channel=ps&ef_id=CjwKCAjwvsvTBhBaEiwAmf-3niL0_BL-F8aPK5N9EXC66KLf4e2JBdor5IkcUSJ2nRSQGOzEy8zoXxoCvWIQAvD_BwE&gads_camp=23527793912&gads_ag=197228065909&gads_ad=813189298265&gads_kw=agentcore&gads_matchtype=p&gads_network=g&gads_device=c&gads_geo=9007584&gad_campaignid=23527793912&gclid=CjwKCAjwvsvTBhBaEiwAmf-3niL0_BL-F8aPK5N9EXC66KLf4e2JBdor5IkcUSJ2nRSQGOzEy8zoXxoCvWIQAvD_BwE. For a deeper walkthrough of AgentCore concepts and best practices, see the https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html to get started.

aws.amazon.com

Amazon Bedrock AgentCore adds memory, policy, and harness in AWS GovCloud (US-West)

Today, Amazon Bedrock AgentCore introduces new capabilities in AWS GovCloud (US-West). AgentCore is the platform to build, connect, and optimize agents. With these capabilities, teams operating in regulated environments can build context-aware agents and take them from prototype to production faster, with the controls needed to scale across their organization.  AgentCore memory gives agents short-term memory for immediate conversation context and long-term memory that extracts persistent insights and preferences across sessions. Agents deliver more intelligent, personalized experiences without teams needing to manage complex memory infrastructure. Policy in AgentCore provides centralized, fine-grained controls for agent-tool interactions that operate outside agent code. Teams author policies in natural language that automatically convert to Cedar, the AWS open-source policy language. Policies attach them to an AgentCore gateway that evaluates each request before allowing or denying tool access.  The managed harness lets developers declare an agent's model, tools, and instructions through configuration and run it in just a few API calls without writing orchestration code, while AgentCore handles the environment, compute, memory, identity, and observability.    To learn more about AgentCore capabilities, visit the AgentCore product page. For a deeper walkthrough of AgentCore concepts and best practices, see the AgentCore Documentation to get started.

ai.warp-studio.com

Amazon Bedrock AgentCoreにおけるAIエージェントの時系列ポリシーによるセキュリティ強化：Dogwoodとゲートウェイ制御の技術的深掘り

Amazon Bedrock AgentCoreでAIエージェントの時系列ポリシーとDogwood言語によるセキュリティ強化を解説。

aws.amazon.com

AgentCore runtime instances are now generally available

Today, AWS announces runtime instances, a new feature in Amazon Bedrock AgentCore that lets you run agents on your own Amazon EC2 instances without managing infrastructure. AgentCore runtime provides purpose-built infrastructure to deploy and operate AI agents securely at scale. Runtime instances complement the existing microVM-based option in AgentCore runtime and give teams running sustained, resource-intensive, or specialized-hardware agents access to the breadth of EC2 instance types, while AgentCore handles provisioning, patching, scaling, and lifecycle management. Using the AWS Management Console, CLI, SDKs, or APIs, you create a capacity provider that specifies the EC2 instance types your agents need, including GPU-accelerated, memory-optimized, and compute-optimized families, and attach your agents to it. Runtime instances support long-running agent sessions of up to 14 days, while the default serverless, microVM-based runtime is designed for sessions of up to 8 hours that need fast startup . You can choose the right compute for each agent, or run a mix, without changing how you deploy or invoke your agents. You can use runtime instances in the following AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Mumbai), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Europe (Frankfurt), and Europe (Ireland). You are charged for the management of the compute provisioned, in addition to your Amazon EC2 costs. To get started, visit AWS News Blog or AgentCore documentation. To learn about pricing, visit AgentCore pricing.

aws.amazon.com

AgentCore runtime instances are now generally available

Today, AWS announces runtime instances, a new feature in Amazon Bedrock AgentCore that lets you run agents on your own Amazon EC2 instances without managing infrastructure. AgentCore runtime provides purpose-built infrastructure to deploy and operate AI agents securely at scale. Runtime instances complement the existing microVM-based option in AgentCore runtime and give teams running sustained, resource-intensive, or specialized-hardware agents access to the breadth of EC2 instance types, while AgentCore handles provisioning, patching, scaling, and lifecycle management. Using the AWS Management Console, CLI, SDKs, or APIs, you create a capacity provider that specifies the EC2 instance types your agents need, including GPU-accelerated, memory-optimized, and compute-optimized families, and attach your agents to it. Runtime instances support long-running agent sessions of up to 14 days, while the default serverless, microVM-based runtime is designed for sessions of up to 8 hours that need fast startup . You can choose the right compute for each agent, or run a mix, without changing how you deploy or invoke your agents. You can use runtime instances in the following AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Mumbai), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Europe (Frankfurt), and Europe (Ireland). You are charged for the management of the compute provisioned, in addition to your Amazon EC2 costs. To get started, visit https://aws.amazon.com/blogs/aws/runtime-instances-persistent-compute-for-production-ai-agents-on-amazon-bedrock-agentcore/ or https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/runtime-instances-how-it-works.html. To learn about pricing, visit https://aws.amazon.com/bedrock/agentcore/pricing/.

aws.amazon.com

Announcing temporal policies and rate limiting in Amazon Bedrock AgentCore

Amazon Bedrock AgentCore announces two new controls: temporal policies for stateful agent authorization and rate limiting for AI traffic. Temporal policies let you define stateful authorization rules that evaluate each request in the context of an agent's prior actions within a session, because a single tool call can be safe in isolation yet harmful given what preceded it. With temporal policies you can enforce workflow sequencing, require that a tool argument exactly matches the output of a prior call, require human approval before taking privileged actions, and enforce data freshness. Rate limiting enables per-user or per-group controls over how much traffic flows to the tools, models, and agents connected to your gateway. Using rules scoped by OAuth or AWS IAM, you can set rate limits on requests across all target types, tokens for inference targets, and concurrent connections to cap long-lived concurrent sessions, aiding downstream service availability and enforcing fair limit distribution.  For regional availability and to learn more, see the documentation, read the announcement blog, and explore the Dogwood reference implementation.

aws.amazon.com

Announcing temporal policies and rate limiting in Amazon Bedrock AgentCore

Amazon Bedrock AgentCore announces two new controls: temporal policies for stateful agent authorization and rate limiting for AI traffic. Temporal policies let you define stateful authorization rules that evaluate each request in the context of an agent's prior actions within a session, because a single tool call can be safe in isolation yet harmful given what preceded it. With temporal policies you can enforce workflow sequencing, require that a tool argument exactly matches the output of a prior call, require human approval before taking privileged actions, and enforce data freshness. Rate limiting enables per-user or per-group controls over how much traffic flows to the tools, models, and agents connected to your gateway. Using rules scoped by OAuth or AWS IAM, you can set rate limits on requests across all target types, tokens for inference targets, and concurrent connections to cap long-lived concurrent sessions, aiding downstream service availability and enforcing fair limit distribution.  For regional availability and to learn more, see the https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html, read the https://aws.amazon.com/blogs/machine-learning/control-agent-behaviors-and-cost-beyond-a-single-action-new-capabilities-in-amazon-bedrock-agentcore/ and explore the https://github.com/dogwood-policy. 

www.synestesia.uk

Read the full article →

aws.amazon.com

Amazon Bedrock launches Web Search for OpenAI GPT models

Today, we are announcing the general availability of Web Search on Amazon Bedrock, a built-in server side tool that performs web search entirely within AWS, enabling OpenAI models (GPT-5.4, GPT-5.5, and GPT-5.6 Sol/Terra/Luna) to ground responses with current web knowledge while maintaining data residency within your secured AWS environment with zero data egress. Previously, adding web grounding required onboarding a third-party search provider, managing separate API keys and billing, building custom orchestration, and conducting additional compliance reviews for each external vendor. Web Search removes this heavy lifting by enable grounding with a single parameter in an existing API call, with no vendor onboarding, no external APIs to orchestrate, and no additional vendor security reviews to conduct. Web Search is built by Amazon, informed by years of experience across Alexa+, Amazon Quick and Kiro. It combines a web index operated by Amazon, spanning tens of billions of documents refreshed continually, with a built-in knowledge graph that provides verified facts. Rather than returning raw pages, Web Search performs semantic snippet extraction, delivering context-efficient results optimized for the model's context window with low latency. Web Search integrates through a standardized tool-use interface, compatible with the OpenAI Responses API. Simply add the web search tool to your API call, and Bedrock handles the entire search lifecycle server-side; a single API call returns a grounded response with citations. Web Search on Amazon Bedrock is generally available today in US East (N. Virginia), US East (Ohio), and US West (Oregon). To get started, read our blog post Introducing Web Search on Amazon Bedrock for foundation model grounding, review the Web Search section in the Amazon Bedrock User Guide for technical documentation, and visit the Amazon Bedrock pricing page for cost details.

aws.amazon.com

Amazon Bedrock launches Web Search for OpenAI GPT models

Today, we are announcing the general availability of Web Search on Amazon Bedrock, a built-in server side tool that performs web search entirely within AWS, enabling OpenAI models (GPT-5.4, GPT-5.5, and GPT-5.6 Sol/Terra/Luna) to ground responses with current web knowledge while maintaining data residency within your secured AWS environment with zero data egress. Previously, adding web grounding required onboarding a third-party search provider, managing separate API keys and billing, building custom orchestration, and conducting additional compliance reviews for each external vendor. Web Search removes this heavy lifting by enable grounding with a single parameter in an existing API call, with no vendor onboarding, no external APIs to orchestrate, and no additional vendor security reviews to conduct. Web Search is built by Amazon, informed by years of experience across Alexa+, Amazon Quick and Kiro. It combines a web index operated by Amazon, spanning tens of billions of documents refreshed continually, with a built-in knowledge graph that provides verified facts. Rather than returning raw pages, Web Search performs semantic snippet extraction, delivering context-efficient results optimized for the model's context window with low latency. Web Search integrates through a standardized tool-use interface, compatible with the OpenAI Responses API. Simply add the web search tool to your API call, and Bedrock handles the entire search lifecycle server-side; a single API call returns a grounded response with citations. Web Search on Amazon Bedrock is generally available today in US East (N. Virginia), US East (Ohio), and US West (Oregon). To get started, read our blog post https://aws.amazon.com/blogs/machine-learning/introducing-web-search-on-amazon-bedrock-for-foundation-model-grounding/, review the https://docs.aws.amazon.com/bedrock/ for technical documentation, and visit the https://aws.amazon.com/bedrock/pricing/ for cost details.