Free to start with credits every month — enough to generate several complete architectures. Pro unlocks all infrastructure code formats (Terraform, Pulumi, Bicep), full security findings with line navigation, unlimited Git and Jira exports, inline code editing, branching for architecture exploration, ZIP and CSV downloads, and credit top-up packs. A full generation (diagram, infrastructure code, documentation, security scan, and cost estimate) costs a few credits. See the pricing section above for full details.

AWS, Azure, and GCP — with hundreds of supported services per provider. Kubernetes ecosystem diagrams and Firebase are also supported. Each provider has dedicated service catalogs covering AI/ML, data engineering, web apps, containers, networking, event-driven architectures, and security.

No. ArchGenie generates the code and exports it via Git PR, ZIP, or inline editor — you run the deployment from your own pipeline with your own credentials and state backend. Cloud credentials and Terraform state are never sent to ArchGenie. The only credentials we hold are the OAuth tokens you authorise for Git providers (GitHub, GitLab, Bitbucket) and Jira, used to push code or create issues.

Generated code, diagrams, and chat history are stored in your account — only you (and your team workspace members, if shared) can read them. We do not use your content to train AI models, and we contractually require the same of our third-party providers. You can delete your account and all associated data from Settings; personal data is removed within 30 days.

The grade reflects the percentage of policy checks the architecture passes after automatic hardening has been applied — A at 90% or higher, B at 80%, C at 70%, D at 60%, F below. The breakdown is per-category (encryption, network, identity, logging, deletion-protection) so you see what drives the score. Lower grades indicate categories that need human judgement: customer-managed keys, IAM scoping, or network isolation choices automation cannot make.

Pricing is sourced from each provider's pricing catalogue. Per-resource costs reflect machine type, instance family, region, and storage tier as declared in the generated code. The estimate covers the resources in the architecture; it does not include cross-region data transfer, cross-cloud egress, or third-party SaaS costs. Where a SKU is missing from the underlying pricing database, the panel marks it explicitly so the gap is visible.

Multi-turn refinement works in the same chat — describe what to change ("add a Redis cache between the app and the database", "switch from RDS to Aurora") and only the parts that depend on the change are re-derived. The rest of the diagram, code, security findings, and cost estimate stay stable. Branching a chat lets you explore alternatives non-destructively.

Yes — the Code Canvas includes a full code editor. Edit any file, save to re-validate, and apply automated fixes across single files or entire projects. Changes propagate automatically across dependent files.

Git export creates a branch and pull request with your infrastructure code files, README documentation, architecture and sequence diagrams, and cost summary. Supports GitHub, GitLab, and Bitbucket. You can also create Jira issues directly from findings.