You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The AgentCore "getting started" instructions are not clear that "local" development actually accesses AWS, and critically apparently provides no indication of which AWS account is being used.
agentcore create --name MyAgent --framework Strands --protocol HTTP --model-provider Bedrock --memory none
cd MyAgent
agentcore status
Status says:
AgentCore Status (target: )
Agents
MyAgent: Local only
Then run:
agentcore dev
Enter a question, e.g.:
What are the benefits of uv over pip?
Expected Behavior
"Local only" means no server access.
If an AWS account is accessed, agentcore status should indicate which account is being accessed.
Actual Behavior
An error occurred, which apparently means that "local only" is not "local only".
Error: An error occurred (AccessDeniedException) when calling the ConverseStream operation: Your account is currently being verified. Verification normally takes less than 2 hours. Until your account is verified, you may not have access to this operation. If you are still receiving this message after more than 2 hours, please let us know by writing to aws-verification@amazon.com. We appreciate your patience.
There is no information about which AWS account is being accessed, and no obvious way to find out.
CLI Version
0.27.0
Operating System
Windows
Additional Context
AWS best practices requires setting up multiple accounts. Users need to be able to know which account is being accessed. This issue is a compounding of the critical oversight of #1727: the --profile switch is not implemented, and it's not even clear if AWS_PROFILE is supported. Therefore, when the "local" setup tries to access something that isn't local, users need to know what it's trying to access.
Improve the documentation to explain that "local" (apparently) only refers to the agent harness; an AWS account is still being accessed.
Description
The AgentCore "getting started" instructions are not clear that "local" development actually accesses AWS, and critically apparently provides no indication of which AWS account is being used.
Steps to Reproduce
Follow the instructions for Get started with the AgentCore CLI:
agentcore create --name MyAgent --framework Strands --protocol HTTP --model-provider Bedrock --memory none cd MyAgent agentcore statusStatus says:
Then run:
Enter a question, e.g.:
Expected Behavior
agentcore statusshould indicate which account is being accessed.Actual Behavior
CLI Version
0.27.0
Operating System
Windows
Additional Context
AWS best practices requires setting up multiple accounts. Users need to be able to know which account is being accessed. This issue is a compounding of the critical oversight of #1727: the
--profileswitch is not implemented, and it's not even clear ifAWS_PROFILEis supported. Therefore, when the "local" setup tries to access something that isn't local, users need to know what it's trying to access.--profileandAWS_PROFILEfrom Support AWS profiles for deployment (e.g.agentcore deploy --profile xxx) #1727.agentcore statusto indicate which profile is being resolved.