
I asked the console why my instance wouldn't connect, and it went and checked
The EC2 Connect tab now runs an agent that checks instance state, IAM, SSM registration and networking for you. Here's what it found on a fresh instance.
I launched an instance, went to connect over SSM Session Manager, and instead of the usual "not a managed node" dead end, the console handed me an agent that actually went and checked.
What it did
I typed one line:
Help me connect to i-0cff63b0edcf8bf4f via SSM Session Manager
And it ran the checks itself, in parallel, not as a checklist for me to work through:


Get help connecting to your instance by using AWS DevOps Agent
-
DescribeInstanceStatus/GetServiceSetting
-
GetConnectionStatus
-
DescribeInstanceInformation
-
ListAttachedRolePolicies
-
DescribeVpcEndpoints
-
DescribeSecurityGroups
-
DescribeRouteTables
That is the exact sequence I would have run by hand. Instance state, then IAM, then agent registration, then network reachability.
It is not just SSM
The "Help me connect with" dropdown covers every path the Connect tab already offers:
- SSM Session Manager
- SSH - SSH client, EC2 Instance Connect
- Instance in private subnet - EC2 Instance Connect Endpoint
So the same agent will diagnose a key pair and security group problem on the SSH path, or an endpoint and IAM policy problem on the EICE path. Each one has its own failure modes and its own set of things people forget, which is exactly the kind of thing worth handing to something that can just go look.
What it found
- Instance running, both status checks passing
- IAM instance profile attached with
AmazonSSMManagedInstanceCore
- All 3 SSM VPC endpoints present and
availablein the right subnet
- Security group
sg-0d2d1b57022db2740with a full outbound allow (0.0.0.0/0)
-
ssm:DescribeInstanceInformationreturned an empty list - not registered as a managed node
- Instance launched ~10 minutes ago (LaunchTime
2026-08-07T07:01:35Z, current2026-08-07T07:11:46Z)



Verdict: nothing is broken. The SSM Agent is still registering.
The part I liked
Two of its checks failed on permissions -
ssm:GetConnectionStatus and the DHMC status. It said so, and then said why it did not matter: the empty managed node list is definitive on its own, and the instance profile with AmazonSSMManagedInstanceCore is present and sufficient.It did not stop on the permission error, and it did not pretend the permission error was the root cause. That is the failure mode I expected and did not get.
The fix it gave
Wait, then verify:
aws ssm describe-instance-information \ --filters "Key=InstanceIds,Values=i-0cff63b0edcf8bf4f" \ --region us-east-1
PingStatus of Online means it is ready. Then:aws ssm start-session \ --target i-0cff63b0edcf8bf4f \ --region us-east-1
Still nothing after 15 minutes - check agent logs via EC2 Serial Console, or reboot to force re-registration:
aws ec2 reboot-instances \ --instance-ids i-0cff63b0edcf8bf4f \ --region us-east-1
SSM Agent registration typically completes within 5-15 minutes of launch.

Where this is going
Connect assist is a small surface, but it is the pattern: the DevOps Agent moving into the operational console paths, not sitting off to the side in a chat panel. The agent had read access, used it, and reasoned about what the gaps in its own access meant.
Usage showed
0.00/-1 hours on this run. Free, for now.I will take a diagnostic that says "this is expected behaviour for a freshly launched instance" over one that hands me a list of nine things to go check.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article