Infrastructure change requests handled by a platform agent
Describe an infrastructure change in Pulumi Cloud and an agent writes the code, previews it and opens a PR for review.
The use case
Platform teams get a steady stream of small requests: add a queue, tag resources, fix a drifted setting. Pulumi's agents work inside Pulumi Cloud with knowledge of your stacks and policies. Given a request, they write or edit the infrastructure code, run a preview, and open a pull request showing the planned changes. This lets developers ask for infrastructure without learning the codebase. Platform engineers still review the PR and preview, check policy results, and approve deployment. Start with low-risk request types such as tagging or adding monitoring, and widen the scope as the team sees how accurate the generated code is.
Example workflow
- Connect the agent to your Pulumi organization and repos
- Submit a change request in plain language
- Agent edits code and runs a preview
- Review the PR and preview output
- Approve and deploy through your pipeline
Tools
- Pulumi
- GitHub
Risks and guardrails
- Agent permissions should not include direct production updates
- Previews can hide replacements that cause downtime; read them
- Policy as code should still run on every change
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Pulumi docs: Pulumi agents: www.pulumi.com/docs/ai/agents
- GitHub MCP server (official page): github.com/github/github-mcp-server (checked 2026-11-09)