Terms of Service.
1. The service
Soverenia Verdict Engine: MCP tool check_package(purl, version, policy?), REST API /v1/check_package, SDK @soverenia/agent-guard and CLI. A verdict is the deterministic, signed output of the engine in the authoritative block (SAFE / UPGRADE / AVOID), HMAC-signed (Ed25519 from M2).
2. Licence
The CLI/Agent Guard is open source (Apache-2.0 at release). The hosted service is licensed per plan: Free (public repos, community support), Team (19 €/developer/month), Enterprise (custom, sovereign deployment).
3. Verdict liability
Verdicts are deterministic outputs of continuously updated public vulnerability and malware intelligence. We do not guarantee that a SAFE verdict means a package is harmless in every context — the verdict covers the data sources at evaluation time. Liability for decisions based on verdicts is limited per the full ToS; professional liability insurance is in place for Enterprise.
4. Acceptable use
No scraping beyond rate limits, no attempts to forge or replay signed responses (anti-replay: timestamp + nonce), no probing of honeypot endpoints. Violations → API key revocation and IP ban.
5. Data processing
For Team/Enterprise, a DPA per Art. 28 GDPR applies (DPA). All data processed EU-only.
6. Governing law
Czech law; disputes with DACH enterprise customers may be agreed individually per contract.