Concepts

These pages explain how AI agents work with Pulumi. To set something up, see Guides. To try it end to end, start with Get started.

Agents and Pulumi

• Agent Skills: Packaged Pulumi workflows that teach coding agents how to work with Pulumi.
• CLI for agents: How the Pulumi CLI is built for agents, from npx pulumi to structured --json output.

Pulumi Neo

• Pulumi Neo: What Neo does, where you can use it, and how much autonomy you give it.
  • Tasks: The unit of work in Neo, including Plan Mode, task modes, and approvals.
  • Permissions model: Which identity Neo acts as on each surface, and how permission and approval modes constrain it.
  • Automations: Scheduled Neo tasks that deliver their results as pull requests.
  • Neo in the CLI: Interactive Neo sessions in your terminal with your local project and credentials.
  • Neo in your editor: Neo inside Zed, JetBrains IDEs, VS Code, and Cursor through the Agent Client Protocol.
  • Pull requests: How Neo opens and updates pull requests through your existing review process.
  • Code reviews: How Neo reviews pull requests against what Pulumi Cloud knows about your infrastructure.
  • Previews: How Neo runs pulumi preview to validate changes before opening a pull request.
  • Instructions: Organization-wide custom instructions and per-repository AGENTS.md instructions.
  • Slash commands: Built-in and custom shortcuts for proven prompts.
  • Model providers: Running Neo with your own model provider credentials.
  • Settings: Access controls, the default task mode, and notifications.
  • Usage limits: Spend limits for your organization and its members.