Amazon adds remote Claude Code support on Bedrock AgentCore with security controls
Amazon’s Machine Learning Blog details a new pattern that lets developers run Claude Code as a remote agent on Amazon Bedrock AgentCore. Each coding task is executed in its own isolated microVM inside the user’s AWS account, preserving the same data‑protection guarantees that Bedrock offers for on‑premise use.
Key points
- Claude Code can run as a remote agent on Amazon Bedrock AgentCore, each task in an isolated microVM.
- The sample creates a GitHub pull request after coding, using short‑lived GitHub App credentials that expire after 8 hours.
- Deploying the solution takes 15–25 minutes, plus 8–12 minutes to register a repository, and incurs a default $200 monthly charge.
The workflow starts with a local Claude Code plugin that forwards a request through an IAM‑signed API Gateway call. A Lambda function screens the prompt with an Amazon Bedrock Guardrails policy, queues the job in SQS and launches an AgentCore runtime session that pulls a container image from Amazon ECR. Inside the microVM the agent clones a trusted GitHub repository, plans changes with Claude Opus, writes code with Claude Sonnet, runs the repository’s own test commands, and finally opens a draft pull request using a short‑lived GitHub App token that expires after eight hours. All network traffic stays within a private VPC, and Mantle’s zero‑operator‑access design ensures AWS staff cannot view prompts or responses.
Deploying the sample stack takes about 15–25 minutes, with an additional 8–12 minutes to register a repository. The solution builds a roughly 1.3 GB ARM64 image and incurs a default $200 monthly charge while idle, covering NAT, VPC endpoints, and related services.
Model page: Claude Code →
Extend Claude Code to remote sessions on Amazon Bedrock AgentCore with security controls
AWS Machine Learning Blog · 6 November 2026
Loading the full article…
This text was published by AWS Machine Learning Blog and written by Andrew (Qingjie) Ren. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- OpenAI, Meta and xAI launch AI agents for long-term task delegation · 8 src
- Study finds AI agents waste 18% of tokens on unnecessary tasks · 1 src
- Amazon Quick defines four agentic primitives based on When, What, Who, and How · 1 src
- AWS shows how to govern AI agent access with Okta and Bedrock AgentCore Gateway · 1 src
- Perplexity introduces Teammates agents for work on its Max plan · 2 src
Comments
via GitHub Discussions