Innovation & Tech

Amazon Bedrock AgentCore adds four lifecycle control points

AWS's Strands-backed AgentCore rollout adds four Lambda hook checkpoints, managed model and prompt defaults, context controls and built-in tools without customer action.

By Amara Osei

4 min read

AWS Adds Strands Defaults and Lifecycle Hooks to AgentCore Harness
AWS Adds Strands Defaults and Lifecycle Hooks to AgentCore Harnesscocoparisienne / pixabay

What's News

  • The Strands-backed AgentCore rollout began for new and existing harnesses on November 11, 2026.
  • AWS announced the upgrade on November 19, 2026, with four deterministic lifecycle checkpoints.
  • Only Lambda hooks can alter the agent loop; Amazon SNS and Amazon EventBridge hooks provide notifications.
  • A support-agent hook denies issue_refund calls above a $500 auto-approval limit.
  • Explicit model IDs, system prompts, truncation policies and allowedTools selectors remain unchanged.

By Amara Osei

Amazon Web Services began rolling out the Strands-backed Amazon Bedrock AgentCore harness on November 11, 2026, adding four deterministic lifecycle checkpoints. AWS announced the upgrade on November 19. The release requires no customer action and adds no new resource type or integration.

AWS says the upgrade moves AgentCore onto Strands, an open-source harness released earlier in 2026 under an Apache 2.0 license. Strands assembles an agent foundation from instructions, tools and context management. AgentCore operates that foundation as a managed AWS service, with each session running in a kernel-isolated microVM.

AWS manages compute, scaling, high availability and tool-execution infrastructure. Customers retain control over models, instructions, tools, permissions and lifecycle policies. The split leaves runtime management with AWS and policy controls with the customer.

What changes under managed defaults?

Under managed defaults, AWS supplies a recommended model that it can silently change as new models are released. It also maintains a task-oriented system prompt and updates it over time. The prompt tells agents to gather relevant context, use appropriate tools, adapt to failures or denied actions, and verify their work before reporting completion.

Automatic prompt caching is initially available for supported Anthropic models through Amazon Bedrock Converse. AWS says it covers the system prompt, tool definitions and conversation history. Repeated calls with the same prompt and tool surface can use lower-priced cache-read tokens, reducing input-processing latency and cost.

AWS’s proactive context management summarizes older conversation turns after context use reaches a configurable threshold, while retaining recent turns. Separate context offloading moves large tool results out of the model’s immediate context. The agent receives a preview and retains access to the complete result when needed.

Which tools does AgentCore now provide?

AWS added built-in tools for planning, delegation, web retrieval and file access.

  • todo_write preserves a structured plan across invocations in the same session, including after summarization.
  • The subagent tool assigns a self-contained task to a fresh worker inside the invocation. The worker shares its limits and tool-call hooks without creating a separately managed harness resource.
  • web_fetch retrieves webpage content. Separate read, write and edit tools let customers grant filesystem permissions individually.

How do the four lifecycle checkpoints work?

Lambda hooks provide the only mechanism for changing the loop. AWS says: “Only Lambda hooks can change the agent loop.” The functions run synchronously and return an allow or deny decision with an optional reason.

Amazon SNS and Amazon EventBridge hooks provide notifications without blocking execution. If a Lambda function times out or returns an error, the configured failureMode applies and defaults to deny. Decisions appear as hookEvent entries in the InvokeHarness response stream.

  • before_invocation runs before the agent processes the request. A Lambda denial stops the invocation with hook_stopped.
  • before_tool_call runs after the model proposes a tool call and before execution. A denial skips the call while the agent loop continues.
  • after_tool_call runs after a tool result exists. A denial retains the result and stops before further inference.
  • after_invocation runs after the invocation completes. A denial reports the decision, but AWS says streamed output cannot be retracted.

AWS’s support-agent example shows the control path in operation. A before_tool_call hook denies an issue_refund call above a $500 auto-approval limit. The harness skips the refund and lets the agent continue so it can report an escalation.

Tool-call hooks also govern subagent calls, preventing delegation from bypassing the policy. Invocation-level hooks run once for the overall invocation.

How does the rollout affect existing configurations?

The Strands-backed rollout reaches both new and existing harnesses. It requires no customer action, and AWS says explicitly supplied configurations remain unchanged. AWS says existing customer Model Context Protocol servers, AgentCore Gateway, Browser, Code Interpreter, memory and customer-provided containers continue to work through their current interfaces.

What keeps an agent’s behavior stable?

AWS’s configuration rules let customers pin three parts of an agent’s behavior. A model ID keeps a specific model. systemPrompt replaces the managed instructions, and truncation retains a chosen context-management policy.

AWS says omitting allowedTools or setting it to "*" accepts its recommended tool set, which can gain capabilities over time. Listing individual selectors maintains a fixed security or capability boundary.

AWS describes explicit configuration as “the control point for workloads that require stable behavior and a defined tool surface.” As managed defaults and recommended tools evolve, customers who omit those settings will inherit AWS’s changes. Pinned models, prompts, context policies and tool selections will remain unchanged under the rollout.

Source: unite.ai

Share this article:

More from Amara Osei

Amara Osei

Show full bio

Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.

2,422 articles

Related articles

« Previous articleNext article »