audytx vs DryRun Security IaC Security

audytx analyzes AWS infrastructure code and Terraform pull requests, presenting security and cost findings in GitHub and to coding agents. Its deterministic, non-AI engine models resource relationships such as invocation topology, encryption chains, network exposure, resource lifetimes, and IAM trust graphs. The maker lists 303 AWS Terraform checks and 27 cross-resource reasoning axes. A GitHub App posts pull-request digests, inline review, and SARIF, using read-only contents permission while writing comments, reviews, statuses, and SARIF uploads. An optional GitHub Actions step uploads a resolved Terraform plan over OIDC without shared secrets. A hosted MCP server exposes the engine to Claude Code, Cursor, and Codex; autofix applies line-anchored changes and rescans for up to three passes. AWS, Terraform, and CloudFormation are supported in stated hosted workflows. Free includes 100 scans monthly; Plus costs $20 a month and Pro $100 a month.

DryRun Security IaC Security scans infrastructure-as-code changes for security risks by analyzing infrastructure and application code together. It covers Terraform, Kubernetes, and other IaC, with checks for overly permissive IAM policies, publicly exposed resources, insecure defaults, and subdomain takeover risks. Scans run in pull requests and provide guidance during infrastructure design. Teams can write natural-language guardrails for public access, permitted cloud regions, approved resource types, and organization-specific requirements; the Custom Policy Agent enforces them on changes. Terraform findings can appear as pull-request comments and Risk Register entries containing the affected resource, risk description, and remediation guidance. Listed integrations include Claude Code, Claude Desktop, Codex, Cursor, GitHub, GitLab, and Slack. The SaaS service supports GitHub.com personal, organization, and Enterprise Cloud repositories, plus GitLab SaaS; custom instances and other source-control systems require contacting DryRun. DryRun says it uses private models, sends no data to public AI systems, stores minimal metadata and findings, and allows code-access revocation through repository permissions. Pricing is on request, aligned with team size, and a free trial is available.