Skip to content
worth noting Tools and apps

AWS introduces V2 platformVersion in Amazon Bedrock AgentCore runtime for faster agent session startup

only one source so far

AWS has described the V2 platformVersion for Amazon Bedrock AgentCore runtime: instead of restarting the container, a session start restores a snapshot of the initialized environment, bringing session startup down to about 2 seconds, with pre-warming under 1 second.

Amazon Bedrock AgentCore runtime runs each AI agent session in an isolated microVM. According to AWS, the new V2 platformVersion changes how a session starts: instead of repeatedly downloading the image and initializing the container for every new session, the environment is prepared and initialized just once, saved as a snapshot, and then restored for each new session. The result, according to AWS, is a faster and more consistent session start (approximately 2 seconds) and lower compute costs, because V2 also releases unused memory.

For agents that need large files, CLI tools, or a browser at startup and cannot keep them in a running process, the AgentCore runtime now supports build hooks — these allow such data to be loaded and prepared already during snapshot creation. For cases requiring a response under one second (typically chat or voice interfaces), AWS describes a pre-warming technique, where a session starts running in the background as soon as the user opens the application or chat window, even before the first request is sent.

The switch to V2 also affects the deployment process: every runtime creation or update now takes several minutes, because this step builds the snapshot. AWS recommends measuring latency using the Latency and OverheadLatency metrics in CloudWatch, to make it clear whether a slowdown is caused by the platform or by the agent's own code.

The source article is a more extensive technical guide and part of it was not available; see the source article for details.

What changed

Why it matters

For companies running AI agents with an interactive interface (chat, voice assistants) on Amazon Bedrock AgentCore runtime, this is an opportunity to reduce the latency of the first response while also lowering compute costs, since there is no need to pay for repeated initialization on every session. For developers, this means having to adjust the code so that all initialization happens before request processing starts, and optionally using build hooks or pre-warming if they need a response under one second.

Two audiences, two different impacts

What this means

01

For individuals

Developers of agents on the AgentCore runtime must move all initialization (imports, clients, opening connections) before app.run(), and for agents that need large files or a CLI, use the newly supported build hooks, otherwise they will not fully benefit from the snapshot advantage.

What to do When migrating an agent to V2, move initialization code before app.run() and consider using build hooks for large files or external binaries.
More practical updates →
02

For a business

Companies running interactive AI agents (chat, voice interfaces) on Amazon Bedrock AgentCore runtime can, by switching to the V2 platformVersion, reduce session startup latency to the order of seconds and cut compute costs by releasing unused memory, at the cost of a longer runtime creation/update process (several minutes to build the snapshot).

Development
What to decide Evaluate whether the company's interactive agents on the AgentCore runtime benefit from switching to V2 and session pre-warming, given latency and cost requirements.
More business impacts →
agents AWS Bedrock optimization serverless Startup

Check the original

Event sources

only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.

1
AWS Machine Learning Blog primary source · first detected Architecture patterns for fast session startup on Amazon Bedrock AgentCore runtime