Pulumi MCP. Audit infra state, deployments, and outputs via chat.

Your AI agent connects directly to your Pulumi infrastructure state, letting you manage cloud resources through natural language commands. You'll treat it like having a dedicated infra ops engineer sitting right next to you.

Stack Management and Organization.
You can list every isolated project environment in the organization using list_stacks, giving you an inventory of all your stacks along with their resource counts and last update times. When you need a fresh, clean sandbox for testing, use create_stack to spin up a new, dedicated stack within a specific Pulumi project.

If that test stack is empty and you're done with it, remember to call delete_stack; this action permanently wipes the environment.

**Metadata Control.**To keep your costs straight and know exactly where something belongs, you can use tags. You retrieve existing key-value labels on any specific stack using list_stack_tags. If you need to add or update a filter—say, marking it as environment=staging or team=backend—you run set_stack_tag.

Reading the State.
When you're unsure what state things are in, start with get_organization. This function pulls high-level metadata about your entire Pulumi organization. To check the specific setup of one environment, use get_stack to pull all the current configuration and details for a named stack. You can also confirm who initiated any change by calling get_current_user, which returns the GitHub identity used for every API call.

Auditing Deployments.
To track what's happened over time, you list everything using list_deployments. This shows you the full history of deployments for a stack, detailing the version number, whether it succeeded or failed, and exactly when those changes were made. To dive deep into one specific run, call get_deployment to get detailed status information for that single deployment version.

Inspecting Live Outputs.
After a successful deploy, you need to know where your stuff lives—the IPs, the URLs, the ARNs. You read all those exported values using get_stack_outputs. This tells you the live configuration data published by your infrastructure program after it successfully deployed.

Mechanics Summary.
When you're dealing with a specific stack, remember these tools: To get its current status and details, run get_stack. To see what was deployed in the past, check the list using list_deployments or focus on one version via get_deployment. If you need to filter resources by metadata, use list_stack_tags or update those filters with set_stack_tag.

You'll never have to open a console just to get these facts.