Agent Versions and Updates

Overview

Monte Carlo publishes new agent images regularly. Releases add features, fix bugs, and patch vulnerabilities in the image's operating system and Python dependencies. Your agent keeps running the version it has until it is upgraded, so a fix reaches it only once the upgrade is applied.

This page covers who triggers an upgrade, how to check which version you are running, and how to tell whether a newer release is available. For the upgrade steps themselves, see Agent Management.

It applies to customer-hosted agents: AWS, Azure, GCP, proxied, and generic platforms such as Kubernetes. The Monte Carlo platform itself is upgraded by Monte Carlo and needs no action from you.

Who upgrades the agent

Monte Carlo does not roll out agent upgrades automatically on a schedule, even where remote upgrades are enabled.

Monte Carlo publishes agent images, scans them for vulnerabilities, patches findings in new releases, and documents what changed in each release.

You track new releases, decide when to apply them, and run the upgrade — or ask Monte Carlo to push one, where remote upgrades are enabled.

Check which version you are running

In the Monte Carlo UI:

1. Open Settings.
2. Select Deployments, under Data and AI.
3. Select your deployment from the list. You can filter by type or status, or search by name or UUID.
4. Read Agent version under Deployment details. The same panel shows the deployment type, agent platform, status, and health.

This value can take 1–2 hours to reflect an upgrade.

With a reachability test: to get the current version immediately rather than waiting for the UI to catch up, open the deployment as above and select Validate, then expand the Connect to agent check. The agent's response reports version: the version it is running right now. See Testing Agent Reachability from Monte Carlo, or use the testAgentReachability API.

Check whether a newer version is available

Monte Carlo does not notify you when a new agent release is published, so check one of the following.

Release notes. For AWS, Azure, GCP, and proxied agents, the apollo-agent releases page lists every version and what changed. Generic agents are built from a separate codebase, so use the hermes-agent releases page instead. You can watch either repository to be notified of new releases.

Image tags. Available tags are published to montecarlodata/agent on Docker Hub. Find the latest version by looking for the tag suffix that matches your deployment:

Each suffix also has a latest tag — latest-cloudrun, latest-generic, and so on — that points at the newest release for that deployment.

If you deployed the AWS Lambda agent with the default Terraform or CloudFormation configuration, it pulls from Monte Carlo's ECR repository rather than Docker Hub, and the tag is the version number alone — 1.13.20, with no suffix.

Monte Carlo CLI. montecarlo agents upgrade --image-tag -1 lists the most recent tags for your platform and lets you pick one.

Compare the version from the previous section against the latest release. If yours is older, upgrade.

Security fixes and vulnerability scans

Monte Carlo scans the agent image daily and patches findings in new releases. Fixes ship forward only: a published image is never rebuilt in place, so an agent running an older tag keeps the vulnerable component until it is upgraded. This is the most common reason a scanner reports a vulnerability that Monte Carlo has already fixed.

If your own scanner flags the agent image:

1. Check the version your agent is running, using one of the methods above.
2. Compare it against the latest release and read the release notes for your deployment, linked above, to see whether the fix is already out.
3. Upgrade the agent, then rescan.
4. If the finding is still present on the latest release, contact Monte Carlo through Bayes or your Monte Carlo representative with the CVE and the component it was matched on.

For how Monte Carlo identifies, prioritizes, and remediates vulnerabilities across its systems, see Security.

Upgrade the agent

See Agent Management for the upgrade steps for each deployment type, including what to change when remote upgrades are disabled.

Related

• Agent Management
• Platform Architecture
• Security