conformis

Conformis is an EU AI Act compliance tool for scanning codebases and generating audit-ready documentation. Its stated workflow covers AI features in a project, classifies their risk under the EU AI Act, and produces documentation in minutes rather than months. The site frames it as a way to ship AI without relying on lawyers.

It supports codebase upload or direct GitHub connection, and it also offers a local CLI mode. The CLI scans a project on the user’s machine and transmits only findings, not source files; the page says it sends file paths, line numbers, library names, a one-line code snippet, and a short description. It also describes --privacy-mode, which strips code snippets client-side, and --dry-run, which prints the JSON payload before anything is sent. The CLI is said to work on laptops, CI runners, and air-gapped machines, and the platform says it runs inside an isolated EU-based container.

The scanning process is described as AST-based analysis for Python, JavaScript, and TypeScript. It identifies imports and inference calls, and the page names OpenAI, Anthropic, Google, HuggingFace, and custom ML models as examples it detects. Detected features are mapped to EU AI Act categories — prohibited, high risk, limited risk, or minimal risk — with plain-English rationale. It generates Article 11 technical documentation with obligations, evidence, and next steps, and the report is exportable as PDF. The page also says scans are audit-ready and GDPR compliant, and that code is deleted after analysis.

Pricing includes a free plan with no credit card required and 3 scans per month, plus a Pro plan at €99 per month billed monthly. The free plan includes compliance score only, ZIP upload, and GitHub integration; the Pro plan adds full gap analysis per article, unlimited scans, PDF compliance reports, remediation guidance, a compliance checklist and executive summary, and priority support. The page also names founders exploring compliance, startups serious about compliance, and teams embedding checks in CI/CD as intended users. It lists Python 3.9+ for the CLI and notes public and private GitHub repositories are supported.