Deployment & infra

Pulumi

Infrastructure as code using real programming languages.

Agents can propose infrastructure as code, run previews, and call Pulumi Cloud APIs with JSON output.

Agent access evidence

CLI

native

CLI supports preview, up, destroy, stack, config, and JSON output.

Evidence

API

native

pulumi api can call Pulumi Cloud REST endpoints non-interactively.

Evidence

Browser

strong

Pulumi Cloud UI supports stacks, previews, resources, and audit review.

Evidence

Account creation

strong

CLI can run locally; Pulumi Cloud self-serve signup.

Evidence

Pricing clarity

strong

Open-source CLI plus public Pulumi Cloud pricing.

Evidence

Docs quality

strong

Docs cover CLI, SDKs, Automation API, and cloud APIs.

Evidence

Sandbox

native

Preview provides a dry-run before infrastructure updates.

Evidence

Readiness score

Strong

Strong: 81/100. Strongest in agent safety; weakest in production reliability.

Machine operability20/25

Agent safety23.2/25

Agent readability17.4/20

Auth and setup12/15

Production reliability8.8/15

Best for

Programmable IaC, previewed infra changes, multi-cloud provisioning, and automation APIs.

Pricing

Open-source CLI plus public Pulumi Cloud pricing.

Auth

Pulumi access tokens, cloud provider credentials, ESC environments, and stack permissions.

Account creation

CLI can run locally; Pulumi Cloud self-serve signup.

Browser support

Pulumi Cloud UI supports stacks, previews, resources, and audit review.

CLI

pulumi

API

https://api.pulumi.com

Limitations

Key limitations: Apply/update operations can change production infrastructure; require review and scoped stack tokens. Deploys, secret changes, IAM, billing, networking, deletes, and rollback paths have high production blast radius; require scoped credentials and review gates. CLI access is powerful but can run destructive local or production commands; prefer dry runs, preview environments, and non-production credentials. API automation needs client-side throttling, idempotency where supported, pagination limits, retries, and audit logging.