• STATUS
  • CHANGES
  • ERRORS
  • HUMAN

How to audit deployment history with your AI

Connect Pulumi to your agent to pull stack updates and deployment logs. Once connected, you can query your infrastructure history using natural language.

Ask AI about this page

Short answer

How can I audit my Pulumi deployments with AI?

Your agent uses the Pulumi MCP to fetch recent stack updates and specific deployment data. You can ask it to identify failed runs or summarize resource changes. This allows you to skip the CLI and get direct answers about your infrastructure state.

The audit outcomes.

What you'll find.

The AI pulls the raw data so you can get straight to the answers.

STATUS

Deployment health

The agent checks recent stack activity to show you which updates succeeded or failed. You'll see a clear list of recent deployment statuses.

CHANGES

Resource diffs

Ask for the specifics of a run to see exactly which resources were modified. The AI identifies the changes made during a specific update.

ERRORS

Failure analysis

When a deployment fails, your agent can pull the specific details for that event. This helps you pinpoint why a stack update didn't complete.

HUMAN

Manual fixes

The AI can read the history but cannot change your infrastructure. You'll need to handle any code fixes or new deployments yourself.

The workflow

What your AI does when the request arrives.

Your agent acts as the interface between your questions and the Pulumi API.

  1. Scan stack history

    The agent looks through your Pulumi account to find recent deployment events.

    list_deployments

  2. Retrieve deployment data

    The agent pulls the full details for a specific stack update you've identified.

    get_deployment

  3. Identify resource changes

    The agent examines the deployment details to see what resources were affected.

    get_deployment

  4. Summarize results

    The agent takes the technical deployment data and explains it in plain English.

    list_deployments

Try it

Copy these to start.

Use these prompts to begin your audit.

Starting points

These are opening lines. Swap out the stack names for your actual Pulumi stack names.

Pulumi Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_v2N5nlYKKcBMg7t6tvswP5zOuY1KsfB14QgLHciO/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Pulumi capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "pulumi-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_v2N5nlYKKcBMg7t6tvswP5zOuY1KsfB14QgLHciO/mcp"
    }
  }
}
Copy into chat04

  • Show me the last five deployments for the production stack.

  • What happened during the most recent deployment in the staging stack?

  • List all recent deployments for my dev stack and tell me if any failed.

  • Get the details for deployment ID 'abc-123' and summarize the resource changes.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Pulumi once, then ask.

Connect your Pulumi account with one click. Your credentials stay encrypted on our side, and you can immediately start querying your stacks through your agent.

Connect Pulumi to your AI

FAQ

How this task behaves.

  • 01

    Can the AI trigger a new deployment?

    No. The AI can only read deployment history and status. You must use the Pulumi CLI or console to run new deployments.

  • 02

    Can the AI modify my infrastructure?

    No. The tools provided only allow the AI to read deployment data. It cannot change, delete, or create resources.

  • 03

    How does the AI know which stack to look at?

    You specify the stack name in your prompt, and the agent uses that to find the right history.

  • 04

    Can I see the resource changes for a specific run?

    Yes. The agent can pull the details for a specific deployment to show you what changed.

  • 05

    Does this work with Cursor or Claude?

    Yes. This MCP works with any compatible client, including Claude, Cursor, and Windsurf.

  • More questions about Pulumi? See everything the Pulumi Connector can do

Connect Pulumi to Claude, Cursor, ChatGPT & more