AWS adds Secrets Manager support to Bedrock AgentCore

Amazon Web Services (AWS) has announced a new capability for the Bedrock AgentCore platform: developers can now connect their own secrets, stored in AWS Secrets Manager, to the AgentCore Identity module, instead of creating separate credentials inside the agent platform itself. This lets companies extend their existing secrets-management processes to AI agents without building a parallel security infrastructure specifically for AgentCore.

How the new integration works

AWS Bedrock AgentCore is a service for deploying and managing autonomous AI agents in the Amazon cloud. AgentCore Identity is the module that handles agent authentication and grants agents access to external systems through so-called credential providers. Before this update, those providers required separate secret configuration inside AgentCore. Now you can point to an existing, pre-configured secret from AWS Secrets Manager, Amazon's service for centralized storage of passwords, API keys and tokens, and AgentCore will reference it directly, without duplicating the data.

What it means for security teams

The main benefit for information security teams is that full control over the secret's lifecycle stays where it was before. The company keeps control of encryption, key rotation, replication, tags and resource access policies, exactly as it would for any other secret in Secrets Manager. This removes a typical risk of adopting new AI services: a sprawl of fragmented credential stores, each with its own access and audit rules, that are hard to control centrally.

Key details of the update:

• Service: AWS Bedrock AgentCore, AgentCore Identity module.
• Source of secrets: AWS Secrets Manager.
• The company retains control over encryption, rotation, replication, tags and resource access policies.
• Using a secret from another AWS account in the same AWS Region is supported.
• Cross-Region sharing of secrets is not supported.
• Secrets connected through Secrets Manager external connectors are supported, meaning integration with third-party secrets managers is also possible.

Why it matters for enterprise AI

For large organizations that have already built governance processes around secrets, with mandatory password rotation, access logging and permissions split by department, being able to connect agents to existing infrastructure without migration is a systemic improvement, not a cosmetic one. Unlike classic applications, AI agents increasingly access sensitive systems such as CRMs, internal APIs and financial services on behalf of a user or an organization. How an agent authenticates, and who controls its credentials, is therefore becoming a key question as agentic AI scales in enterprise environments.

The ban on cross-Region use of secrets reflects AWS's typical approach of isolating data by geographic region. Companies operating in several jurisdictions will have to configure secrets separately in each Region where their agents run, which may require extra planning for multi-Region AgentCore deployments.

Secrets management is especially acute for AI agents because, unlike static services, an agent can independently initiate many calls to different external systems within a single session, which increases the number of points where an access key could be compromised. Being able to plug in a long-tested secrets rotation and audit process, rather than creating a new isolated store just for agents, reduces the so-called attack surface and makes it easier to meet internal security policies and the external regulatory requirements that apply to large organizations in the financial, healthcare and government sectors. The update also reflects a broader AWS trend: gradually turning the scattered services of the Bedrock platform into a single ecosystem integrated with the familiar AWS infrastructure tools that companies have used for years, whether or not they run AI agents or still work with classic cloud services.