Pulumi shipped 17 release events in 30 days, and the interesting part is not raw activity. It is where that activity points. The May launch is Pulumi repositioning infrastructure as something both humans and agents operate through the same CLI, state model, and review loop.

ToolVitals scores Pulumi at 100 for health, 100 for shipping, and 100 overall. The repo has 25,245 GitHub stars, 20 GitHub releases in 90 days, and an Apache-2.0 license signal, so ToolVitals classifies it as OSI-approved open source.

The signal: Pulumi is building for agents without abandoning operators

The official site now leads with “infrastructure as code for humans and agents.” That is not subtle positioning. Pulumi still sells the familiar pitch, infrastructure as code in TypeScript, Python, Go, C#, Java, and YAML, but the new emphasis is on agents doing useful infrastructure work inside the same system operators already trust.

The strongest proof is the cluster of May releases. Pulumi announced pulumi do for direct create, read, update, delete, and query operations across Pulumi-supported providers without a project, code, or state. That is a sharp turn toward imperative cloud operations, but it is still tied to Pulumi providers and resource models rather than a separate toy CLI.

Neo got the same treatment. Pulumi moved Neo into the terminal with pulumi neo, added GitHub and Slack entry points, launched MCP and cloud CLI integrations, and added scheduled automations that can open pull requests. That is a coherent bet: agents should investigate, edit, preview, and propose infrastructure changes where teams already work, not in a separate chat box with no audit trail.

The bet is governance, not magic

The better story here is not “AI writes infra.” Everyone is saying that now. Pulumi’s angle is that agents need boring controls: RBAC, audit logs, read-only modes, approval modes, pulumi preview, pull requests, and stack state.

That matters because infrastructure agents are dangerous if they skip the normal change path. Pulumi’s recent posts keep routing agent work back through previews, PRs, scoped integrations, and Pulumi Cloud identity. Even pulumi do, the most ad-hoc feature in the batch, is framed as a quick path for one-off resource operations with a route back to full IaC when the resource becomes permanent.

This is why the 17 release events matter. They are not random churn. They show a product team turning the CLI, cloud console, collaboration surfaces, and integration layer toward the same agentic infrastructure thesis.

What ToolVitals cannot infer

ToolVitals can say Pulumi is active, highly scored, and shipping frequently. It can also say the public repo and first-party announcements support the current agent-focused positioning.

ToolVitals cannot tell you whether Neo produces good pull requests, whether pulumi do fits your internal controls, whether users like the new CLI paths, or whether Pulumi Cloud pricing works for your team. It also cannot measure code quality from stars and release counts alone. Those require hands-on evaluation.

The GitHub signal is still strong. The repository page showed recent activity around v3.243.0 and public repo positioning as “Infrastructure as Code in any programming language.” But ToolVitals metrics remain the source of truth for the counts in this post.

Comparisons: Pulumi versus nearby tools

Pulumi’s 221.6 hot score puts it below n8n at 240.0, but that comparison needs care. n8n has 189,586 stars and 24 release events in 30 days, but ToolVitals classifies it as fair-code, not OSI-approved open source. Pulumi has fewer stars at 25,245, but it carries an Apache-2.0 OSI-approved open-source signal.

Against Jenkins, the comparison is tighter. Jenkins has 25,290 stars, almost identical to Pulumi’s 25,245, but Pulumi logged 17 release events in 30 days versus Jenkins at 10. Both have shipping scores of 100. Pulumi’s recent public narrative is simply more concentrated around agents and CLI-driven infrastructure work.

SST is another useful DevOps comparison. SST has 26,014 stars and 11 release events in 30 days with a 94 shipping score. Pulumi has slightly fewer stars, more release events, and a 100 shipping score. That does not make Pulumi “better,” but it does show stronger current release intensity in ToolVitals’ data.

Recommendation

If your platform team already uses Pulumi, evaluate the May 2026 agent features now, especially pulumi neo, GitHub and Slack workflows, scheduled automations, and pulumi do. The reason is concrete: Pulumi is shipping the agent layer inside the existing Pulumi workflow, not beside it.

If your team is choosing an IaC platform from scratch, treat Pulumi as the agent-forward option to test first. Run a real trial: one provider update PR, one drift investigation, one incident-style lookup, and one direct pulumi do operation. If those fit your approval model, Pulumi’s current shipping pace is a real advantage.

Sources